What's New
Latest platform improvements
The brand plan is $99 a month
A brand that signs up directly now pays $99 a month for the full audit across five engines, every page the engines point to read and checked, five prompts checked every day and the fixes ready to paste in. The Terms of Service dated 11-14-2026 state the new price.
How it works
New direct sign-ups are billed at $99 from today. Nothing changes for agencies: Starter, Growth, Scale and the per-client bands (the plan that bills each client on its own, at a price that falls as you add clients) are as before, and the price an agency charges its own clients is still its own.
Where to find it
Pricing page, and Terms of Service section 5.1
Why it matters
One price that reads as the full product it is, and the same number as the highest per-client price, so an agency never has to explain why its vendor sells cheaper to its own client.
Your referral link and your agreements, in the portal
Your referral code and link, the accounts you referred and your monthly statements now live in your own portal, next to a page where you can read every agreement you accepted.
How it works
Open Agency in the sidebar and pick Referrals to see your code, copy your link and follow each referred account through its statements. The figures are the ones the partner portal shows. Pick Agreements to see which document and which version you accepted, on what date and by whom, and to open the full text of each one. If a newer version of the Terms or the Agency Agreement has been published since you accepted, both texts are there to read.
Where to find it
Agency (Account for a brand), then Referrals and Agreements
Why it matters
You no longer need a separate sign-in to check what a referral earned, and you can reread the exact words you agreed to without asking us for a copy.
The report reads front to back
The audit report now runs in four chapters, with three charts and four appendices, so a reader can follow it from the first page to the last. The web page, the PDF, the Word file and the Markdown file all use the same order.
How it works
Chapter one is the summary: your score, what it means and the engine scorecard. Chapter two gives each engine one card with its scores and an excerpt of what it said about you. Chapter three holds the evidence: how you compare with the competitors you named, the sources the engines used and Unprompted Visibility. Chapter four is what to do, ending with the roadmap and the method. Three charts show each engine's overall score, how many questions each engine named you on, and who the engines name instead of you. Every question is listed once, by engine, in Appendix A. The full engine answers are in Appendix B, the full competitor answers in Appendix C, and the raw data in Appendix D, which prints in the Word and Markdown files only.
Where to find it
Any audit: open it and read the report, or download it as PDF, Word or Markdown.
Why it matters
You can hand a client the first chapter and know it stands on its own, and when they ask where an answer came from, the full text sits in one appendix and not scattered across the cards.
Download a lead's report as a PDF
The report a free-check prospect received can now be saved as a PDF from the lead's card. It prints in your agency's own report template, so you can attach it to an e-mail you send from your own address.
How it works
Open Pipeline, then Leads, open the lead and click Download PDF. The file holds the same report you see under View report: your agency's template when you have one, and the platform's default template when you do not. It is named after the prospect's brand and the date you downloaded it, which keeps a folder of attachments easy to sort.
Where to find it
Pipeline, then Leads, then a lead's card
Why it matters
When you send the report yourself, the conversation with the prospect stays in your inbox and under your name from the first reply onward, which is where it belongs when the next step is a call with your team.
API keys for your own tools
You can now create an API key and use it to read your clients, audits, reports and monitoring data from your own scripts and tools. Keys are the first step toward connecting Sextant AEO to Zapier and Claude, which are still to come.
How it works
Open Settings, then API keys, name the key and create it. You can then read your clients, audits, reports and monitoring data from your own scripts and tools. The key appears once, with a copy button and the line "Store it now; it is not shown again." With it you can list and read your clients, list and read audits, download a report as PDF, DOCX or Markdown, and read your monitoring brands, prompts, runs, metrics, competitors, sources, engine set and alerts. Three things can be written: you can create a client, run an audit and add a tracked prompt. Everything else on the API stays closed to a key. A key acts as the person who made it, so it carries that person's permissions and sees only the clients that person sees. Only people with permission to manage settings can create or revoke keys, and you can revoke one from the same page whenever you like. Whoever writes the script puts the key in the Authorization: Bearer header of a request to https://app.sextantaeo.com/api/v1/me, which answers with your agency's name and the key's name. The same header works on every call listed here.
Where to find it
Settings, then API keys
Why it matters
If you keep a spreadsheet or a dashboard of your own, a key lets it pull your audit scores and monitoring figures on a schedule, so the report you send a client and the sheet you keep for yourself read from the same source.
The PDF reports no longer split a score block, a card or a chart across two pages
Every printed report (the audit, the progress report and the kickoff plan) now breaks pages only between whole things. A score tile, a score row, a finding, a callout box, an engine scorecard and a chart print on one page, a heading never sits alone at the bottom of a page, and a long table repeats its header row on every page it runs onto.
How it works
The three documents share one set of print rules that keep each block whole and keep every heading with what follows it. A table that is longer than a page still runs over, row by row with no row cut in half, and its header row prints again at the top of the continuation. A table that is shorter than a page no longer gets pushed to a fresh sheet, which is what used to leave a heading alone on a near-blank page.
What changed
Before: The Dimension Breakdown section could print four of its score rows on one page and the fifth on the next, an engine heading could end a page with its scorecard on the following one, and the question grid started on a fresh sheet after a page that held only its title.
Now: The five score rows print together, each engine heading shares a page with its scorecard, and the question grid starts under its title and carries its header row onto every page it runs onto.
Where to find it
Any report downloaded as a PDF: Client, then Reports, then Download PDF, and the Kickoff plan PDF on the Plan tab.
Why it matters
The PDF is the thing an agency forwards to its client and the thing a prospect judges the platform on. A chart or a scorecard cut in half reads as unfinished, whatever the numbers say.
The menu stays open, the plan and its recommendations point at each other, and you can send a client their intake
Three asks from an agency this week, shipped the same day. The left menu now opens by default instead of showing icons only. A recommendation says which plan step it is and jumps to it, and a plan step jumps back to its recommendation, with the row you land on outlined. And a client's Portal row has a Send intake button that e-mails them a sign-in link straight to the questions.
How it works
The menu opens on every screen; a small arrow at its top collapses it, and the choice is remembered on that device. On the Recommendations tab, each card reads In the plan: Weeks 5 to 8, step 4 of 16, with the due date, and that line is a link to the step. On the Plan tab, a step whose recommendation is loaded shows Open in Recommendations beside its status. Either jump outlines the row or card it lands on for a few seconds. On the Portal tab, Send intake sits between Preview and Resend link; it sends the viewer a branded e-mail whose button signs them in and opens the intake page, and the link expires in 24 hours like any sign-in link.
Where to find it
The menu on every page. Client, then Recommendations, and Client, then Plan, for the links. Client, then Portal, for Send intake on an active viewer's row.
Why it matters
An agency told us the menu was hard to find, that moving between a recommendation and its place in the plan took guesswork, and that there was no way to ask a client for their intake without writing the e-mail by hand.
A night an engine was down now shows a gap in its chart
When an engine answers nothing on a night, because its vendor is down or its account could not pay, that engine's monitoring chart now shows a gap for the night instead of a 0 % visibility that looked like a collapse.
How it works
The moment an engine fails because our account with it has run out of credit, we stop watching that engine until it answers again, and our team is told within the minute. While it is stopped we record nothing for it, so the chart shows a gap and not a zero. Audits keep running during a pause, and the report gate holds an audit that is missing an engine, so a report with a gap in it waits for us before it reaches your client.
Where to find it
The monitoring charts, where the engine's line shows a gap for the night.
Why it matters
A false zero in front of a client costs you more than a missing night does, because it reads as a brand that dropped out of an engine overnight. The gap reads as what it is, a night we got no answer from that vendor.
See how we check a report before your client does
Every check we run on a report is now written out in plain words on a public page, so you can read what a report is held for before you forward one to a client. The page also counts the reports we have checked and the ones we held, once there is enough history to count honestly.
How it works
The page lists all 12 checks in two groups. The first group reads the scores and decides whether the whole report waits, and the second reads one section at a time and holds only that section. Each check is written as one or two plain sentences, and a check we are still testing says so on its line. A closing section says what the checks do not cover, which is what the engines said about your client, because the report quotes those answers as they came. Once the count is two weeks old and at least one report has been held, the page shows how many reports were checked and held since 1 November 2026, how many failures were fixed automatically, how many a person released after reading them and how many were fixed before release, and the median hours from hold to release, with a table for each check.
Where to find it
sextantaeo.com/checks, linked from the site footer and the methodology page. The help article Why a report says Being checked explains a hold, and the audit page links to it under a held section.
Why it matters
You send these reports to your own clients, and a line that says Being checked raises a question you should be able to answer without writing to us. The page lets you say what was checked, what a hold means and what happens next. It also lets you show a prospect that a report is read before it is released, with our own counts beside the claim.
Your drafts suggest only places that fit your business
The places your drafts suggest for a mention now fit the kind of business your client runs, and a pitch that cannot stand on its facts is left out rather than printed with a gap. When a question in the audit asked for a fact we could not find, the draft marks it for you to fill in.
How it works
Before a report publishes, each draft is read against your client's record. A review e-mail only names sites where customers leave reviews. When an audit question asked for a fact, like the founding year, and neither the site nor the intake answers it, the draft carries a Fill-in line so you know what to add. If one engine answered fewer than half of its questions while another answered at least half of its own, we ask the first one again before your scores publish. That gives it a fair chance to be measured instead of left out.
Where to find it
The drafts in the audit report and every export, and the held sections listed on the audit page.
Why it matters
A draft that sends your client to a competitor's blog, or asks customers for reviews on a site that takes none, costs you credibility the first time your client tries it. We would rather leave a draft out than hand you one you have to apologize for. When a section is held because your client's record was wrong, correcting the record is the fix. Once we mark the hold as fixed by that change, the check runs again within the hour and the section clears if it passes.
A report waits until it passes our checks
Before a report reaches you, the audit now checks its own answers and sections. A part that fails is held with a one-line reason in its place, and the scores wait when the answers themselves are in doubt.
How it works
The checks run after the engines answer and the fixes are drafted. If the answers about your brand do not add up, the audit stays in progress and the progress screen says why. If one section fails a check, the scores publish and that section shows the reason until a person has looked at it, which usually takes under a day.
Where to find it
The audit progress screen, the audit page, and the report in every format.
Why it matters
Two reports last week quoted an engine that admitted it did not know the brand, at full marks. The checks exist so that a report you forward to a client has already been read once.
Reports print every question and page title in full
The Where You Show Up grid used to print long questions cut to 80 characters with an ellipsis, and each engine's card listed at most eight questions. Both now print in full, and page titles in the technical tables are no longer cut at 80 characters.
How it works
A long question wraps onto a second or third line inside its column in the grid. Each engine's card lists every question it named your client on and every question where it did not, with the count in the heading, such as Questions it named you on (15).
Where to find it
The audit report, its PDF, the Markdown and Word exports, and each engine's card. Page titles appear in the tables of pages the engines cite.
Why it matters
A reviewer reading a report could not see which questions were asked, because the open questions were cut off in the grid. A client judging an audit needs the exact questions the engines were asked.
The report's first page now reads from the audit's own data
The box after the cover and the Top 3 Findings are now worked out from each audit's own numbers, so a high score can no longer sit above a page showing the engines rarely name the brand. Each answer is quoted in full once, and an answer about a different company is shown only when no other exists, with a label.
How it works
The box says what the composite measures, which is how the engines answer questions that already name the brand, and puts beside it how often they named the brand when asked open questions about its category. It names any engine that described a different company or said it did not know the brand. A green score reads as a strong position only when the brand is named in at least one open answer in five and every engine knows it, and an audit that asked no open question says discovery was not measured. Top 3 Findings put that identity problem first when the brand is named in fewer than one open answer in ten, then rank the dimensions by the points each one costs rather than by percentage, and suggest only the schema types the client's kind of business uses, the same ones the technical table prints. The technical section asks a publication for NewsMediaOrganization and Article rather than Service, a software product for SoftwareApplication and a local business for LocalBusiness, and a Site issues section lists broken basics on the site itself, such as a navigation link that leads to a missing page, a robots.txt that answers with a web page, a sitemap that leaves out the homepage and Privacy or Terms links that go nowhere. A written fix that needs a fact the audit could not confirm opens with a Fill in line naming it. An answer that describes a different company or says the engine does not know the brand is shown only when no other answer exists, with a caption that says so, and later sections point back to an answer instead of printing it again.
Where to find it
The first two pages of the audit report and its PDF, the Markdown and Word exports, and each engine's card, which now also lists the questions where its answers did not name the client. The Site issues section sits after the Technical Gap Analysis.
Why it matters
Two reports checked line by line before sending called a brand well positioned while their own Unprompted Visibility page showed it named in a handful of open answers, and one showcased an engine describing a different company. A client who catches that once stops trusting the rest of the report. The verdict and the findings on the first page now rest on numbers printed in the same report.
More answers that cannot place your client now count as not knowing it, and answers about a different business are flagged
From today an audit recognizes more of the ways an engine says it cannot place your client, and those answers earn no Brand Recognition, Presence Quality and Share of Voice points. A second model also checks whether each answer describes a different business that shares your client's name, and the report's Misidentification box lists any engine it catches.
How it works
Earlier versions already caught many of these admissions, such as "not familiar with" or "no publicly available information", but they knew the "not a company I can verify" form only with the word company. Scorer version 5 adds platform, tool, product, vendor and publication to it, along with an engine asking which of several similarly named businesses you mean, and an answer like that is scored the way every disclaimer is, with no Brand Recognition, Presence Quality and Share of Voice points, Market Position as computed and the lower Brand Sentiment points a disclaimer keeps. The second model reads every other answer that names your client and decides whether it describes your client, a different business with the same or a similar name, or neither, judged against the category and one-line description in the client's intake (or the client's industry when the intake is empty). For now its finding is recorded and listed while the answer's scores stay as they were. Once we have checked it against answers we have labeled by hand, an answer it is at least 80% confident about will be scored like a disclaimer too, and a new entry here will say so.
What changed
Before: An answer in which the engine said your client was not a platform it had reliable information about could still earn all 20 Brand Recognition points, because that admission was only recognized when the engine said company.
Now: The same answer earns no Brand Recognition, Presence Quality and Share of Voice points, and its Brand Sentiment is the disclaimer's.
Where to find it
The scores of every audit run from today, in the report and the PDF, on the audit page in the agency app and in the client portal, where the audit page's details show Scorer version 5. The report's Misidentification box also lists an engine the second model caught.
Why it matters
A score built on answers that never knew your client tells them they are visible when the buyers asking about them hear that the engine cannot place them. The lower number is the honest one, and the Misidentification box shows which engine to work on first.
Platform recommendations now come from what the engines cite
The review and directory sites an audit recommends are now the ones the engines actually cite when they answer your client's own questions, minus the listings the client already has. Each recommendation says why that site was picked.
How it works
The audit counts how often the engines cite each review or directory site across every answer it collected. A site needs at least 3 citations to count, and up to 3 sites get a profile draft, the most cited first. A site you listed under Review profiles in the intake is never drafted again, and its address goes into the Organization schema. When the engines cite that site but not your client's own page there, you get a Complete your profile card for the listing instead. If an audit has no answers to read, the sites come from a short list for the client's industry, and the line under the title says so. If the engines cite no review or directory site at all, the audit tells you that in place of the drafts, because a listing nobody cites would not move the answers.
Where to find it
On each recommendation card and in the report's Third-party profiles section, with a line under the title saying why that site was chosen. The same line sits under the matching step in the plan, in the app, in the client portal and in the kickoff PDF. In the intake, the Review profiles field now says what the list is used for.
Why it matters
An agency's own audit told it to claim a profile it had held for three weeks, and a marketing agency was sent to doctor-review sites. A recommendation that asks for work already done, or points at the wrong kind of site, spends the trust the rest of the audit earned. Every pick now rests on what the engines cite, and it shows its count.
Every audit now says how stable its own answers were
Each engine's result in an audit now carries a label, Decision-grade or Directional, set by how often three answers to the same question agreed on whether your brand was named. The free check says it is directional, because it asks each question once.
How it works
An audit asks every question three times on every engine. A question earns full marks when all the answers agree, half marks when all but one do, and none otherwise, and an engine's agreement is the average over its questions. An engine is Decision-grade when that average is 75% or better across at least 6 questions, and Directional below that. The audit also gets one overall label, worked out from every question on every engine together. Agreement measures consistency, not visibility: an engine that never names you three times out of three is perfectly consistent and scores 0 on recognition.
Where to find it
The scorecard and each engine's findings in the report and the PDF, the audit page in the agency app, the client portal's dashboard beside the score, and the free check's result.
Why it matters
A score from one answer is a single draw. The label tells you and your client how far to lean on a number before acting on it. The research page already publishes this arithmetic for its ten public questions, and every audit now carries the same figure for its own questions.
Your market, in the kickoff plan
The kickoff plan PDF now carries a Your market page written for the client. It says who they are, sets out the three tiers of competitors with a line on why each one competes, and names where they win and where they face pressure.
How it works
The page comes from the client's Market brief, so it appears once a brief exists and is left out when there is none. It opens with one sentence on who the client is, built from the category, audience and place the brief found, then shows each tier as a table with the competitor's name, where it is based and why it competes. Two short lists follow, one for where the client wins and one for where it faces pressure, and a last line gives the date of the research. The page speaks to the client as you and sits after the introduction and before the three windows, so the rest of the plan reads as it did. The picks you make in the interview decide which competitors get tracked. The page shows the whole market the brief found, whatever you picked.
Where to find it
Clients, a client's Plan tab, Download kickoff PDF. On the Competitors tab, once a research run has finished, Review these in the interview opens the guided interview at the competitor step with the brief's tiers listed and tier 1 ticked, so you can pick who to track and carry on to the prompts.
Why it matters
A client who sees the market the way the engines and the web describe it understands why the plan tracks the competitors it does. The plan stops asking them to take the list on trust, and you have the research in the document you already hand over.
Monitoring prompts proposed from the Market brief
The Market brief now proposes the prompts to monitor. You get one comparison for each competitor you track, one category question for each of the client's service pillars, and local versions of those questions when the client serves a territory, and each prompt says where it came from.
How it works
When a brief is written, during a brand's first audit or when you press Research the market, Sextant turns it into suggested prompts. No model is asked, so the same brief always gives the same prompts. A comparison reads the way a buyer would ask it, for example Acme vs Rival Agency, which is the better choice for growth-stage software companies, and there is one for each competitor you track, each pending suggestion from the web and each tier 1 competitor in the brief. A category question asks who the best providers of one service pillar are, for the audience the brief names. A client with a city, state or region also gets those questions asked locally, while a client that serves the whole country or the world does not, because a local question there would only repeat the category one. Every prompt carries a line saying where it came from, such as Comparison with Rival Agency (tier 1: Direct regional competitors), or the service pillar it answers and the territory it is local to. Nothing is tracked until you accept it. On the Prompts tab, Generate now starts from the brief's prompts and lets the AI fill what is left of the batch.
Where to find it
The onboarding interview's prompt step, where each suggested prompt shows its origin beside it, and a client's Monitoring set-up, which lists the suggestions the same way. The Suggested prompts panel on a client's Prompts tab shows the same line under each prompt.
Why it matters
Prompts that name the real competitors are the ones whose answers move when the work lands. Starting from the brief means you begin with those prompts, with the reason for each one on the page, instead of writing them one at a time.
The activity feed no longer says credits were used up when today's check has not run yet
A day with no monitoring check on record read as Monitoring skipped: plan credits used up, even when the only reason was that the day's check had not happened yet. The feed now says Today's check has not run yet for today and No check recorded for this day for a past day, and keeps the credits wording for a check the plan's credits really stopped.
How it works
The daily check writes a record for each client it reaches, and a check that stops because the monthly credits are spent writes one too. The feed used to treat any day without a record as a credits problem. It now reads the date: today without a record means the check is still to come, and a past day without a record means the check did not reach the client that day. Only a record that says the credits ran out produces the credits line.
Where to find it
Clients, a client's Activity tab, and the same feed in the client portal's dashboard.
Why it matters
An agency reading the feed before the evening check saw a credits warning that was not true, and a client reading their portal saw the same line. The feed now says what it knows and nothing more.
Each plan step opens into the work behind it
A step in the plan now opens into the recommendation it ships, with its artifact and deploy note, or into the finding or outreach target it was built from.
How it works
Open a client's Plan tab and look under any step. A step that ships a recommendation has Show the recommendation, which opens the same card you see on the Recommendations tab, while Measure Impact and Mark shipped stay on the step itself. A step that came from a failed site check has Show what was found, which gives the check, its status, why it matters and the page it points at, and a step that came from a page the engines cite has Show the target, which gives the site, an example page, the action to take, the engines that cite it and the competitors the audit found named there. The link runs the other way too. Each recommendation card now says where it sits, for example Plan: window 2, step 4 of 16, due 14 November, and that line takes you to the step.
Where to find it
Clients, a client's Plan tab, under each step. On the Recommendations tab, the plan line sits under the title of each card.
Why it matters
Until now a plan step and a recommendation were two lists you matched by title, and you had to assume which part of the plan a fix belonged to. Now you can open a step and see the work behind it, or open a recommendation and see which step owns it and when it is due.
A market brief for each client, from two research engines
Research the market now writes a market brief for the client: who the client is, up to three tiers of competitors with a line on why each competes, where the client wins, where it faces pressure, and the sources. You choose from it which competitors to track.
How it works
Perplexity and Gemini each search the live web with the same question. They work from the client's name, website, industry and location, from the category, one-liner and services you entered in Intake, and from the text of the client's homepage. Claude then merges their two answers into one brief, and every competitor shows which engine named it. A client with a territory gets direct regional competitors, multi-regional competitors and specialists in its core service, and a product gets its category, the incumbents that sell it as a module and the adjacent platforms. Review sites, directories, social networks and AI assistants are left out, and so is the client. If one engine fails, the brief comes from the other and says so. The research runs in the background, so you can leave the tab while it works. It shows which step it is on, how long it has been running and a rough time left, and the brief appears when it finishes. Track on a competitor's row puts it in Suggested competitors as a pending suggestion tagged From the web, and from there you Accept or Dismiss it as you always did. A new brand also gets its brief during the first audit. The wait after the audit now shows the same steps, reading the site, researching the market, writing the brief and proposing competitors and prompts, and the set-up opens with the brief's competitors listed above the names the engines gave, the first group already ticked. The category, audience, services and service area it found are offered in the questionnaire, tagged From the market brief, for you to keep or change.
Where to find it
Clients, a client's Competitors tab, in the Market brief panel above Suggested competitors. The onboarding wizard's competitor step offers Research the market when a client has no brief yet, and every client's first audit writes one.
Why it matters
The old Research button gave you a list of names with no reasons attached. The brief gives you the reasoning an analyst would write down, so you can see why each company belongs on the list before you track it, and you have something concrete to show the client. It also keeps the market's picks apart from the engines' picks, because the two lists answer different questions.
Inactive clients are tagged in your client list
A client you have set inactive now carries an Inactive tag on its card, so the list shows at a glance which clients you are working and which are parked.
How it works
An inactive client does not count toward your plan's client limit, and it cannot be audited until you reactivate it, which needs a free slot on the plan.
Where to find it
Clients, on the card of any client you have set inactive. The Deactivate and Reactivate control on the client page is unchanged. The count in the top right of your dashboard now reads live clients against your plan, with the inactive count beside it.
Why it matters
A roster with parked clients on it used to read as one list of equals, and a report run on a parked client went out looking current. The tag and the audit refusal keep the list and the reports honest without taking anything away from you.
The robots.txt fix now explains the three kinds of AI crawler
The robots.txt fix in Fix-It now comes in three labeled sections, and the audit reports the agents that read a page when a person asks about your client in a list of their own.
How it works
Each section of the fix opens with one line saying what it is. Search agents, such as OAI-SearchBot and PerplexityBot, index a page so an engine can cite it, and the first section allows the ones the audit found blocked. User fetchers, such as ChatGPT-User, Perplexity-User and Claude-User, read a page when a person asks an engine about your client, and the second section allows those. Training crawlers, such as GPTBot and ClaudeBot, collect pages for future models, and the third section leaves the decision to your client: a block there keeps the brand out of future training data and costs nothing in search. The audit treats a blocked user fetcher as a visibility problem, the same as a blocked search agent, and the AI crawler access finding names user fetchers in a sentence of their own. The older anthropic-ai token is now checked as a training crawler too.
What changed
Before: The fix was one block of Allow lines under a single comment, and ChatGPT-User, Perplexity-User and Claude-User were counted with the search agents.
Now: The fix has a section for search agents, one for user fetchers and one for training crawlers, and the audit reports each in its own list.
Where to find it
On a client's Recommendations tab, open the robots.txt recommendation and read its How to deploy text. The AI crawler access row of the Technical Findings section of the audit report names the same three classes.
Why it matters
Pasting the fix used to leave one question open, which is whether it also lets AI companies train on the site. It does not, because the third section adds no live rule, and the block now says so where the client reads it.
Reports in your brand's look
The audit report and the progress report now use the same design as the kickoff plan PDF, with your agency's two colors as the accents.
How it works
All three documents are now built from one stylesheet. Your primary color sets the rule above each section, the headings and the edge of each callout, and your secondary color sets the score numbers, the top edge of the score cards and the line in the trend charts. Where you have not set a color, the document prints in the platform's default blue and gold. The type is one face throughout, with the numbers lined up in columns, so a table of scores reads straight down.
What changed
Before: The progress report printed on cream with terracotta and navy, and the audit report in navy and blue, whatever colors you had set in Agency Settings.
Now: Both read like the kickoff plan: a light page, one typeface, and your primary and secondary colors as the accents.
Where to find it
Open an audit report from a client's page, or download a progress report PDF from the client's Progress tab. The audit report's on-screen view uses the same design as its PDF.
Why it matters
A client who gets the kickoff plan, the audit and the progress report should see one set of documents from one agency, and until now the last two looked like they came from a different product.
Brands accept the Terms only at activation
A direct brand now accepts the Terms of Service when it activates its subscription, and the Agency Agreement is for agencies.
How it works
The activation page of a brand account shows the Terms of Service and one checkbox that reads I accept the Terms of Service. The Agency Agreement card is no longer on that page, and Sextant records a single acceptance for the brand, the Terms, with the version and the time. An agency sees the page it saw before, with the Terms and the Agency Agreement together, because the Agency Agreement covers reselling and white-labeling the platform and a brand does neither.
What changed
Before: A brand ticked one box for the Terms of Service and the Agency Agreement, and both acceptances were recorded.
Now: A brand ticks one box for the Terms of Service, and only that acceptance is recorded.
Where to find it
The activation page of a brand account, which opens before the first payment.
Why it matters
A brand that buys a subscription is a customer of Sextant, and the Terms are its contract, so a resale agreement was a document it had no reason to read or sign.
Every plan step now says what to do
Each step in the AI Visibility Plan now carries a What to do line, so your team and your client can act on a step without opening anything else first.
How it works
Every step reads an instruction written for its kind of fix: in your agency's words on the Plan tab, and in your client's own on the portal and in the kickoff PDF the portal offers. A step that ships a Fix-It recommendation links back to it, and when the recommendation's deployment note is a full description, the step adds the first three sentences of that note after the instruction. A step that comes from a failed technical check fills in the page and links to the help article when one exists. A step about a cited page says how many answers cite it without naming your client, which of your competitors the page names when the page check read it, and the pitch page the publisher itself offers when the audit found one. The first sentence of each follows the steps listed under Next up in the Monday digest and the Progress Report.
What changed
Before: A step showed its title, its page, its engines and a checkpoint, and a step from a technical check said nothing about how to fix it.
Now: Every step has a What to do line under its title, and a linked step opens its recommendation from that line.
Where to find it
A client's Plan tab, the Plan page in your client's portal (where an outreach step also links the cited page), and the kickoff PDF you download from the Plan tab.
Why it matters
Someone has to know what a step asks for before it can be done, and until now that meant opening the recommendation or asking the agency. The instruction now sits on the step, worded for whoever reads it.
Export the Prompts and Sources tables as CSV
The Prompts tab, the Sources tab and each prompt's drill-down now have an Export CSV button, so you can hand a client's marketing team the same tables you see on screen.
How it works
Click Export CSV and your browser saves a spreadsheet file named after the client, the table and the date. The Prompts file has one row per tracked prompt with its topic, tags, status, country, engines, responses, visibility, citations, sentiment and last run date, over the same 30-day window the tab shows. The Sources file has one row per cited domain, or one row per cited page when the toggle is on URL, with the citation count, the mover status and the first and last dates seen, and at the URL level it also says whether the page names your client and which competitors it names. The drill-down file has one row per engine answer in the window you picked, with the answer text, the brands it mentioned and the sources it cited. It holds the 200 most recent answers, and when a window has more than that, the last row says so.
What changed
Before: The Prompts and Sources tables could be read on screen and nowhere else.
Now: Each table, and each prompt's drill-down, downloads as a CSV file that opens in Excel or Google Sheets.
Where to find it
Open a client, choose Monitoring, then Prompts or Sources. The Export CSV button sits in the toolbar above the table, next to the Domain and URL toggle on Sources. Open a prompt and the same button is at the top of its drill-down panel.
Why it matters
A client's marketing team usually wants the tracked prompts and the cited sources in a file they can sort and filter on their own, and until now the screen was the only place they lived.
Updated Privacy Policy
We have published a Privacy Policy and Terms dated 10-25-2026; the de-identified-data wording in both takes effect 11-08-2026 and account holders were e-mailed. It now names what our free AI visibility check and our This week in AEO newsletter collect, how long we keep it, and where to send a security report.
How it works
Section 2 lists what the free check collects: the website and brand name you enter, your e-mail address, your name if you give it, and the IP address and browser user agent of the request. The check runs only after you click the link we e-mail you. The same section says the newsletter keeps your e-mail address and the time you confirmed the subscription. Section 7 adds how long each is kept: an unconfirmed free-check request is deleted after 7 days, a confirmed check loses its IP address and user agent after 12 months, and a newsletter subscription lasts until you unsubscribe or ask us to remove you. Section 6 now describes our de-identified analytics store more carefully, and section 8 sends security reports to [email protected].
What changed
Before: The policy did not mention the free check or the newsletter, said the de-identified store cannot be used to re-identify a customer or Brand, and sent security reports to [email protected].
Now: The policy names both, says the store is designed so that it cannot reasonably be used to re-identify a customer or Brand and that we remove any record found to fall short of that, and sends security reports to [email protected].
Where to find it
The Privacy Policy page on the Sextant website, sextantaeo.com/privacy.
Why it matters
When a client asks what Sextant collects about them, or about the people who try the free check, the answer is written down in one place along with how long each item is kept.
Free checks run after the requester confirms their e-mail
Your free visibility check now e-mails the person who filled it in a link to confirm, and runs only after they click it.
How it works
A prospect enters their website and work e-mail as before. We send one e-mail in your branding to that address, with a single button and no score. The button opens a page on which the check starts, and the prospect sees the results there and gets the report by e-mail as before. The link works for 24 hours. Submitting the form again for the same address sends the same link again, at most once every 10 minutes.
What changed
Before: The check ran the moment the form was sent, and the report went to whatever address was typed, whether or not its owner asked for it.
Now: The check runs only once the owner of the address clicks the link, so the report only goes to someone who asked for it.
Where to find it
Nothing to set up. In Pipeline, then Leads, a submission waiting for its click shows as Awaiting confirmation under its own filter; confirmed checks show as Confirmed.
Why it matters
Nobody can use your form to send your branded e-mail to a stranger, and every lead in your pipeline is an address whose owner asked for the check.
Dates now read MM-DD-YYYY everywhere
Every date Sextant shows you or your clients is now written month, day and year with dashes, such as 10-22-2026, on every page, e-mail, report and PDF.
How it works
One rule prints every date: MM-DD-YYYY, with a leading zero on a one-digit month or day. Only the way a date is written changed. In the portals, a time still shows in your own time zone and in the style it had before, after the date, and the date is your own day for that moment. E-mails, PDFs and reports have no time zone of their own, so where they show a time, such as the moment a trial ends, it is in UTC and says so.
What changed
Before: Dates came in several styles: "Oct 22, 2026" in the portals, "October 22, 2026" on reports and PDFs, "22 October 2026" in e-mails and on our website, and the browser's own style on some tables.
Now: Every surface writes the same day the same way: 10-22-2026.
For example
An audit completed today reads 10-22-2026 on the client's dashboard, in the AI Visibility Progress Report, on the kickoff PDF and in the Monday digest. A trial that ends on a New York evening reads as the next day in UTC, for example 11-10-2026 at 01:30 UTC.
Why it matters
A date in a report you send a client is written the same way as the dates in their portal and in their inbox, so nobody has to work out whether two styles name the same day.
The free check explains its five dimensions, and Brand Sentiment stops penalizing plain descriptions
The free visibility check now names each of the five dimensions under the composite and says in one line what it measures, and from today Brand Sentiment takes points away for negative framing while a neutral, accurate description of a brand keeps most of them.
How it works
Each dimension on the free-check result reads as its score out of its maximum, with one line on what it measures in the same words the audit report uses, and the composite stays the headline above them. For Brand Sentiment, an answer that names a brand the engine recognizes keeps all 40 of the 40 points when its framing is positive, 34 when it is neutral or factual and 16 when it is negative, and a mixed answer lands between those according to how its praise and criticism balance. An answer in which the engine says it does not know the brand is scored the way it was before, and the scorer now recognizes more ways an engine says so, so a brand nobody recognizes gains nothing. The agency and client overviews now name the five dimensions the way the reports do, in title case, so the verdict and the score cards say Brand Sentiment where they said Sentiment and Share of Voice where they said Share of voice.
What changed
Before: The free check drew five bars under the composite, each labeled with the dimension's internal name and its score over its maximum, with nothing on what a dimension measures, and a neutral description of a well-known brand kept 22 of the 40 Brand Sentiment points.
Now: Each of the five reads as its proper name, its score out of its maximum and one line on what it measures, and the same neutral description keeps 34.
Where to find it
The result of the free visibility check, on the Sextant AEO site and on any check your agency embeds or links that does not use its own report template. The e-mailed report and the lead in Pipeline, then Leads, already showed the five scores and still do; a lead from before today keeps the sentiment wording it was first sent with.
Why it matters
A prospect can see which part of the score is low and what it measures, which gives your first conversation something specific to start from, and brands the engines describe accurately no longer read weaker than they are.
Safer site crawls, and a check that says when it could not finish
When Sextant crawls a client's site, every connection now goes only to that site's own verified address, and the JavaScript-free content check says Not fully measured when the crawl could not see everything a browser would.
How it works
Before a crawl starts we look up the site's address once, confirm that it is a public one, and from then on connect to that address and nothing else, on the standard web ports 80 and 443. The browser we render pages in no longer loads files from other hosts, such as scripts from a CDN, web fonts or analytics tags, so content that only those scripts would add never appears on the page we read. When the crawl blocked any of those scripts, the JavaScript-free content row reads Not fully measured and says how many requests to how many other hosts were blocked, and that row counts as neither a pass nor a gap. The pages that cite your client are read under the same rule, on ports 80 and 443 only.
What changed
Before: A crawl that could not load a site's off-site scripts could still report that AI crawlers see what visitors see, and a website address typed as a bare name such as acme, with no domain ending, was accepted and only failed when the audit ran.
Now: The row reads Not fully measured with the number of blocked requests and hosts, and a bare name such as acme is refused as soon as it is entered.
Where to find it
The Technical Findings table of every audit report, in the web view, the PDF and the Markdown and Word downloads, on the JavaScript-free content row.
Why it matters
A report you send a client never says a check passed when the crawl could not finish it, and the crawler can only ever reach the public site you asked us to audit.
The AI Visibility Plan: twelve weeks of fixes from every audit
Every completed audit now writes a twelve-week plan for that client: the gaps the audit found, put in order and split into three four-week windows, with an owner and a due date on every step.
How it works
Each step comes from the audit: a Fix-It recommendation, a technical finding the site failed, or a page the AI engines cite that leaves your client out. Weeks 1β4 hold the foundations, such as crawler access and schema, Weeks 5β8 the content, and Weeks 9β12 earning mentions. Inside each window the step that pays back most comes first, in this order: engine access, then citations, schema, person and author pages, earned content, case study and reviews. Every step names the page, the engines and the score dimension it targets, the artifact to use, and whether your agency or your client owns it. You set the start date and each step's owner and due date. A step ticks off when its recommendation is shipped or measured, and a re-audit writes a new version of the plan that keeps every finished step ticked.
What changed
Before: The audit report's roadmap sorted fixes into Now, Then and Later, and the order lived only in the report.
Now: The roadmap column of the audit report reads the plan's windows, Weeks 1β4, Weeks 5β8 and Weeks 9β12, so a fix sits in the same weeks in the report, the plan and the kickoff PDF.
Where to find it
The Plan tab on each client, between Recommendations and Insights, where Download kickoff PDF prints a copy in your branding for the client. Your client reads the plan under Plan in their portal. Its progress also appears in the AI Visibility Progress Report and, for clients with monitoring, in the Monday digest. For an audit that finished before this release, press Build the plan on the tab. A direct brand's home opens on its plan once the plan has steps, and every client's Overview and your client's portal dashboard start with Plan at a glance: the progress, the next steps due and a link to the plan.
Why it matters
Your client gets a dated list of what ships when and watches it tick off between audits, and your team gets one order of work that the report, the portal and the PDF all agree on.
Daily monitoring add-on withdrawn
We have taken the daily monitoring add-on off sale. Buying it did not make monitoring run more often, and it was never charged to anyone.
How it works
The add-on was meant to move a client from the standard monitoring allowance to a daily one. Nothing in Sextant made that happen when it was bought, so the client's checks would have run on the same schedule as before. We have removed it from the pricing page, the help center and our billing catalog. Monitoring itself has not changed: a direct brand's tracked prompts are checked daily, and an agency client's monthly credit allowance is sized for about three checks a week.
What changed
Before: The pricing page and the help center offered daily monitoring of the full prompt set as an add-on.
Now: Nothing is sold as an add-on. Each plan monitors what the plan itself includes.
Why it matters
Anything you pay for should change what Sextant does for you. This did not, so we stopped selling it.
Referral partner program
Apply at sextantaeo.com/partners to earn a share of Net Profit for every agency or brand you introduce to Sextant AEO.
How it works
Apply through the form at sextantaeo.com/partners. Sextant sends you the agreement when it accepts your application. Once you accept it, you get a link that carries your code. Sharing it sets a 90-day cookie on the visitor, so an agency or brand that signs up from that browser within the window is credited to you, as long as it was not already talking to Sextant. Introducing someone in person or by e-mail instead? Send [email protected] the company, the contact and the date before they sign up, and we pre-register the introduction. On the Standard schedule, a referred agency earns you 30% of its Net Profit in its first Commission Year, stepping down to 10% from year 4 on, and a referred brand earns a flat 15% for as long as it pays. Statements are issued monthly, and you sign in to see them by magic link at partners.sextantaeo.com.
Where to find it
sextantaeo.com/partners to apply, and partners.sextantaeo.com to sign in once accepted.
Why it matters
Anyone already recommending tools to agencies and brands can turn that into recurring income, and Sextant handles the tracking, the statements and the payments.
Updated Terms and Agency Agreement
We have published the Terms of Service dated 10-20-2026 and Agency Agreement version 1.1. No client is free on any plan, plan changes are self-serve and prorated, and the Terms state the direct brand price of $49.95 a month.
How it works
The Terms set out the published prices: the direct brand plan at $49.95 a month, the three agency bundles, and per-client billing priced at the band your client count falls in. No client is free on any plan. Plan changes are self-serve. Switch tiers any time under Account, Manage billing: the change takes effect immediately and is prorated. On per-client billing, your fees follow the client count you set there. New sign-ups accept these versions. If your agency accepted an earlier Agency Agreement, it stays on that version until you accept version 1.1, and we write to you separately about what changes and when.
What changed
Before: The Terms quoted an old direct brand price, and the Agency Agreement promised a free first client and tier changes on request at the next billing date.
Now: Both documents state the current prices, that every client is paid for, and that a plan change applies at once, prorated.
Where to find it
The Terms of Service page on the Sextant website. For a copy of the Agency Agreement, e-mail [email protected].
Why it matters
The documents you accept say what the product does: what each plan costs, that every client counts, and that you change plan yourself.
Per-client billing is back, with no free client
An agency can again sign up to pay per active client instead of buying a bundle. Every client is paid for, the first one included.
How it works
At sign-up, or when you activate an existing account, choose Per active client and the number of clients to pay for, one at the least. The price per client falls in steps as your total grows, and every client is billed at the band your total falls in. When you reach the number you pay for, adding or bringing back a client is refused with a note that points to Account, Manage billing, where you raise the count. Switch tiers any time under Account, Manage billing: the change takes effect immediately and is prorated.
What changed
Before: From 19 October an agency had to start on Starter, Growth or Scale.
Now: An agency starts on a bundle or pays per active client, and no client is free on either.
Where to find it
The Per active client band on the pricing page, the sign-up form, and the plan choice when you activate an account.
Why it matters
An agency between bundle sizes, or above the largest bundle, pays for the number of clients it chooses rather than a whole bundle.
Demo viewer: a read-only login to show prospects
A new built-in role, Demo viewer, gives you a login to put in front of a prospect: it shows the clients you pick, read-only. Contact details, private intake answers, leads, team e-mails, notes, audit logs and billing are hidden; names published on the client's own website stay.
How it works
Invite a teammate with the Demo viewer role and tick the clients it may show. That login sees scores, audits, reports, monitoring and recommendations for those clients, under a "Demo view" banner, and every change it tries is refused. Our servers take out what could reach a person before anything reaches the screen: intake interview notes and contact entries, e-mail addresses, phone numbers and postal addresses in the audit and its Fix-It schema, who made each change, alert recipients, your notes on recommendations and the penetration tester's name read "Redacted for demo" or come back empty. Team, Roles, Settings, billing, Leads, Free-audit setup, the audit log and every export (CSV, the client data export, and the Word and Markdown report downloads) are not there at all; the report stays as a web page and a PDF.
What changed
Before: Showing a prospect your Sextant meant a Client Viewer login, which still showed client contacts, interview notes and portal users.
Now: A Demo viewer login shows the work and hides the ways to reach the people behind it.
Where to find it
Agency, then Team: "+ Invite teammate" and choose Demo viewer. Agency, then Roles lists it with the other built-in roles.
Why it matters
A live demo on your own clients is the strongest pitch you have, and this one shows a prospect the work without handing over a client's contact details or letting anyone change a setting while they watch.
Agencies start on a bundle; the per-client plan is retired
From 10-19-2026, a new agency account starts on Starter, Growth or Scale. Paying per client is no longer a way to sign up, and no client is free.
How it works
Sign-up and in-app activation offer the three bundles: Starter covers 5 clients, Growth 15 and Scale 40, and every client counts toward the bundle. When a bundle is full, adding or bringing back a client is refused with a note that names the tier with room. Switch tiers any time under Account, Manage billing: the change takes effect immediately and is prorated.
What changed
Before: An agency could pay per active client instead of buying a bundle, and its first client was free.
Now: Every agency is on a bundle, and every client counts toward it.
Where to find it
The agency plans on the pricing page, the sign-up form, and the plan choice when you activate an account.
Why it matters
Your bill is one bundle price, and the number of clients it covers is printed next to it, so there is nothing to count client by client.
A Back button in the guided set-up, and list answers that save as you add them
Every screen of the set-up after the first now has a Back button, the browser's back and forward arrows move between the screens you have visited, and list answers save a moment after you add or remove an item.
How it works
Back shows the screen before, with what you saved there. Change an earlier screen and press Save and continue, and that screen is saved and the next one opens; your place in the set-up stays where you had reached. Each screen has its own address, so the browser's arrows step through the screens you have visited in this visit. In a list, such as your services, your proof or the questions people ask before they find you, an item is saved a moment after you add or remove it and a quiet Saved shows next to the list. Single answers, such as the year you were founded, still save with Save and continue. If the browser's back arrow takes a brand account out of the set-up to its home, a card there says which step you were on, and Resume opens that step.
What changed
Before: The browser's back arrow left the set-up for the page before it, which looked like starting over, and items added to a list were lost if you left the screen without saving.
Now: Back and the browser's arrows move one screen at a time, list items are kept a moment after you add them, and the brand home brings you back to the step you were on.
Where to find it
Every screen of the brand and agency set-up except the first and the closing one; the Resume card on a brand account's home while the set-up is unfinished.
Why it matters
You can check or correct an earlier answer without losing your place, and a slip of the back arrow no longer costs you the answers you had just added.
ChatGPT answers in audits now come from GPT-5.4
From 10-19-2026, the ChatGPT engine in every audit and in monitoring asks OpenAI's GPT-5.4 model instead of GPT-5.5.
How it works
Each audit asks ChatGPT the same questions as before, through the same scoring, and only the model that answers has changed. Claude, Perplexity, Gemini and Grok are unchanged. Every answer records the model that gave it.
What changed
Before: ChatGPT answers came from GPT-5.5.
Now: They come from GPT-5.4, which costs half as much per token.
Where to find it
The raw data appendix of the Word and Markdown report downloads lists the model behind every answer.
Why it matters
ChatGPT is by far the most expensive engine we ask, and moving it is how we keep your price where it is.
The questions step now suggests answers from your audit
The "What people ask before they find you" card in the setup questionnaire now offers answers drawn from the audit, the same way the company and positioning fields already did.
How it works
When the audit finishes, we read what the AI engines and your site say about you and suggest the uses, problems, jobs, issues and symptoms a customer would mention. A suggestion we are confident about is filled in, tagged with where it came from (for example "Named by ChatGPT and Gemini" or "From your site"); you keep or remove each item. One we are less sure about shows grayed out under its list until you press "Use this".
What changed
Before: The five lists on that card started empty, so you had to think of every answer yourself.
Now: They open with what the audit found, and you edit from there.
Where to find it
Step 6 of the setup walk, the questionnaire section of "Run the interview" on a client's Intake tab, and the intake page in the brand portal.
Why it matters
Each item becomes a question we can ask the engines for you, so filled lists mean a deeper audit with less typing.
Your first audit now shows what it is doing
The first-audit step of sign-up and of adding a client now shows which stage the audit is in, how many engine answers are in, how long it has run and about how long is left.
How it works
An audit asks each of the five engines every question in its plan, several times over, then reads your site and drafts your fixes, then finishes the report. While the engines answer, a bar counts the answers in against the number the audit planned when it started. The step shows the time since you started the audit and an estimate that adjusts to how fast this audit is actually going, so a slow run does not claim to be almost done. Reading the pages the engines cite starts once the audit is finished.
What changed
Before: One stage was lit for the whole run, which could look stuck for ten minutes or more.
Now: The stage, the answers counted in, the elapsed time and the time left update every few seconds.
Where to find it
Sign-up for a brand β Your first audit; adding a client in the agency walk β the audit step.
Why it matters
The first audit is the longest wait in getting started. Seeing it move tells you it is working, and the note says you can leave the page: the results appear on your dashboard when it finishes.
Signing out now ends your session on the server
Signing out of the agency portal or the brand portal now cancels the session on our side, not only in your browser.
How it works
A session is two parts: a short-lived token your browser uses for a quarter of an hour and a longer-lived one it uses to stay signed in for a week. Signing out already canceled the longer-lived one when the browser sent it. It now also cancels every live one for your account when the browser cannot name its own session, and a brand portal sign-out cancels that session straight away instead of letting it run to its expiry.
What changed
Before: A copy of your sign-in token made before you signed out could keep working until it expired.
Now: Signing out cancels it on the server.
Why it matters
If you ever suspect a device or a shared computer, signing out is now a real way to shut that access down.
Your account pages now speak brand
If you signed up as a brand, Settings and Team now open as Account settings and People, with only the fields a brand account uses.
How it works
Account settings keeps your brand name, your locked Sextant address, your monitoring usage and the Your data and Close account cards. The logo, color, email sender and custom domain fields are for agencies that resell Sextant under their own name, so they no longer appear on a brand account, and saving only updates your name. People lists everyone who can sign in to your account. Every role you can give covers your whole brand, so there is no client access list to fill in when you invite someone or change their role. The agreement you accept at checkout is headed Brand agreement.
What changed
Before: A brand account saw Agency Settings with branding and custom domain fields, and a Team page that offered client-by-client access.
Now: It sees Account settings with its name, and People with roles that cover the whole brand.
Where to find it
Account in the sidebar: Settings opens Account settings, Team opens People.
Why it matters
The pages you use to run your account now describe your account, with nothing to fill in that does not apply to you.
Close your account and download your data
Account now has Close account, Keep my account and Download my data, with a read-only banner that shows exactly when your data is deleted.
How it works
Close account, in the Account menu, asks for your password and a confirmation, then schedules your subscription to end on your next billing date. From then your account is read-only for 30 days: sign-in still works, reports and billing history stay visible, but nothing new can be audited or edited. A banner shows your closing, read-only or deletion date throughout that window. Keep my account, in the same menu, cancels a scheduled close at any point before it takes effect. After closing you can still reactivate through checkout at any time before deletion, which keeps every report and audit you already have.
Where to find it
Account menu in the agency-admin portal, visible while open and while read-only.
Why it matters
Closing, downloading your data and reactivating are now things an agency does for itself, on its own schedule, instead of emailing support and waiting; every date that matters is shown in the product, not only in the Terms.
Guided client onboarding for agencies
A new agency account now opens into a five-screen set-up that ends with its first client added, audited and interviewed. Every client page now has a Run the interview button for the clients that follow.
How it works
The set-up starts with the agency's name, logo and primary color, a screen you can leave for later. Next you add the first client: type its name and website, and Sextant reads the site for industry, location and competitors for you to check. The client's first audit starts as soon as it is added. The interview screen then asks the same questions a brand answers for itself, phrased as what to ask the client on a call, with a notes box for each section and a button that invites the client to their portal. A closing screen points to your next client and to inviting a teammate.
Where to find it
Automatic on a new agency account's first sign-in. For any later client, open the client, go to the Intake tab and press Run the interview.
Why it matters
Every client starts with an intake taken from a real conversation, so the fixes and reports that follow are built on what the client told you rather than on guesses.
Guided set-up for new accounts
A brand account now opens into a nine-screen guided set-up built around its first audit; an agency account opens into a five-screen set-up built around adding its first client. Both can be left and resumed at any time.
How it works
The brand walk runs the first audit, then builds the brand's story screen by screen: how AI describes it, what it does, who it competes with, its proof points, the rest of the intake questionnaire, the prompts people ask AI about it, and alerts, portal and settings. The agency walk sets the agency's own name, logo and color, adds the first client with industry, location and competitors read straight from its site, runs that client's first audit, then interviews for the client the same way the brand walk does. Every screen saves on its own, so closing the tab and signing back in later resumes on the exact screen you left. Three screens on the brand walk, and the agency name step, can be skipped and finished later.
Where to find it
Automatic on first sign-in after checkout or activation. A finish-later card on Portfolio or the brand home links back to it until it is done.
Why it matters
A new account gets a working first audit and a usable client record in minutes, without an empty dashboard or a support ticket to figure out what to do first. The agency's interview screen is also the standard way to onboard every client after the first one, from that client's own Intake tab.
Sign up and start a free trial without waiting for us
Brands and agencies now pick a plan on the pricing page, enter a card in Stripe Checkout, and get a live account on the spot with a seven-day free trial. Nobody provisions anything by hand.
How it works
Sign-up creates the workspace, the first admin and, for a brand, the first client. Stripe confirms the card, the account opens, and a brand's first audit starts at once. Billing state now controls access: a paused account cannot sign in, and a past-due one shows a banner with an Update card button.
Where to find it
sextantaeo.com/pricing, then Start free trial. Billing lives under Account, Manage billing.
Why it matters
An agency can put a prospect on the platform during the call that sold it, and a brand can be running its first audit two minutes after deciding to try.
A public help center, What's new and the roadmap are live
The help center at sextantaeo.com/help explains how Sextant measures, what every number means, how to run an agency on it, billing, and security, in 60 articles across 11 categories, with search and a seven-lesson Start here course.
How it works
One question per article, written from what the product renders today, with a formula or a worked example from our own audit wherever a number appears. What's new and the roadmap are now public pages too.
Where to find it
sextantaeo.com/help, sextantaeo.com/whats-new, sextantaeo.com/roadmap, and Help in the site navigation.
Why it matters
Your clients and prospects can read how the measurement works without asking you, and a procurement reviewer can read the security section before you are asked.
Sextant AEO is open to the public
From 9 October the platform you run is publicly available: a public site with published pricing ($199 direct, agency tiers from $349), methodology and a request-access form, published Terms of Service and Privacy Policy, and a one-business-day provisioning promise. Nothing changes inside your workspace.
How it works
sextantaeo.com now carries the pricing page, the methodology page and a request-access form for agencies and for brands that come direct. Every account is still set up by hand: we provision a new agency or client within one business day of the request. The Terms of Service and Privacy Policy are published at sextantaeo.com/terms and sextantaeo.com/privacy and apply to every account from 9 October. Support stays [email protected], answered within one business day.
What changed
Before: Access by invitation only; pricing and terms shared by hand.
Now: Public pricing, published terms, a request-access path, and the same white-label workspace you already run.
Where to find it
sextantaeo.com for the public pages, with Terms and Privacy linked in its footer. Your app address, your clients' portals and your reports are unchanged.
Why it matters
Prospects and clients can now read what Sextant is, what it costs and what we promise under our name, while everything they see from you stays under yours.
Engine outages are our problem, not an alert in your inbox
When an AI engine stops answering for a day (an expired vendor key, a provider outage), Sextant's operators are paged. Agencies and clients no longer receive that alert, see it in the alert history, or find it in the weekly digest.
How it works
The daily monitoring pass still checks every engine for every client. If one returns nothing usable all day, that now goes to Sextant support only, once per engine per day, with the vendor's error attached. Your alert rules are the three you can act on: visibility drop, share-of-voice overtake and citation lost.
What changed
Before: An engine outage emailed every agency admin, listed the vendor's raw error in Alerts, and showed an unlabeled fourth rule card.
Now: Nothing changes on your side; we fix the engine and your next run carries on.
Where to find it
Client β Alerts shows the three rules and only alerts you can act on.
Why it matters
A working platform is Sextant's responsibility, and an alert you cannot act on is noise.
Agency accounts sign in at their agency address only
An agency login now works only at that agency's own address, such as youragency.sextantaeo.com, and the sign-in page tells you where to go if you land somewhere else.
How it works
Every agency has its own address on the platform. Signing in with an agency account at the platform address, or at another agency's address, is refused with a short message naming the right one. Existing sessions are not affected.
What changed
Before: An agency account could sign in at any platform address and only hit errors later.
Now: It signs in at its own agency address, and the page points you there if you mistype.
Where to find it
The sign-in page at your agency address. Bookmark it; the magic-link emails already use it.
Why it matters
Each agency's workspace stays its own, and a stray session can't wander into the wrong one.
Change your password from inside the app
An Account menu now sits in the top-right of the admin, with Change password and Sign out.
How it works
Open Account in the header, pick Change password, enter your current password and the new one. The change applies to your login only; it does not touch your teammates or your clients. Forgot it entirely? The login page's "Forgot your password?" link still sends a reset email.
What changed
Before: The change-password page existed but nothing in the app linked to it, so most people never found it.
Now: It is one click from every page, next to Sign out.
Where to find it
Top-right of every admin page β Account β Change password.
Why it matters
Rotating a password should take ten seconds, not a support request.
Fix-It artifacts and prompt suggestions now come from Claude Opus 5.5
The model that drafts your Fix-It artifacts and monitoring prompt suggestions moved from Claude Opus 5 to Claude Opus 5.5, the newer release in the same tier. Same artifact types, same review flow; every artifact records which model wrote it.
How it works
Every Fix-It artifact and every prompt suggestion is drafted by a Claude model working from the audit findings and the client's intake profile. That model is now Opus 5.5. Audit scoring is untouched: the five engines we query to measure visibility, including the Claude engine, are exactly as before, so scores stay comparable.
What changed
Before: Artifacts and suggestions were drafted by Claude Opus 5.
Now: They are drafted by Claude Opus 5.5, checked side by side on a live client before the switch.
For example
Re-running Fix-It on a client produces the same set of artifact types as before; the FAQ questions and profile drafts are drawn from the client's verified site facts and intake answers, as they were.
Where to find it
Client β Recommendations (every newly generated artifact), and Monitoring β Prompts β Suggest prompts. Existing artifacts are not regenerated.
Why it matters
Keeping the writer on the current model is the cheapest way to keep artifact quality up, and this release costs less per artifact than the last one.
A new look, and a clearer story on every page
Every page now opens with the verdict and what to do about it, in your agency's colors.
How it works
The score is still the score. What changed is what sits next to it: a one-sentence verdict, the three actions for the week ranked from the pages the engines trust, the fixes already drafted and the alerts that fired, and the proof β which pages the AI reads about you and what it took from them. Your agency's logo and color now carry through the admin as well as the client portal.
What changed
Before: A composite number, four boxes of internals and tables to read.
Now: The verdict, the plan, the proof, then the movement β in that order.
Where to find it
Portfolio home, every client's Overview, every audit page, and the client portal.
Why it matters
A client should understand where they stand in ten seconds, and what you're doing about it in twenty.
Monitoring sentiment is now judged, not scored
Every brand mention in monitoring is now classified positive, neutral or negative by a decision model instead of being given a number by a language model, and a page that merely lists a brand now counts as neutral.
How it works
For each brand mentioned in an answer we ask a decision model one question β does this answer recommend, merely list, or criticize the brand β and it returns probabilities for the three. The stored score is the probability of positive minus the probability of negative, on the same β1 to +1 scale as before, so trends and reports read the same way.
What changed
Before: A brand that was simply listed among others often scored mildly positive, so sentiment ran warm.
Now: Listings count as neutral; only genuine praise or criticism moves the number.
For example
A managed-security client that appeared in twelve 'top MSSP' lists and one critical review used to show a warm positive trend; it now shows neutral with one negative day β which is what the answers actually say.
Where to find it
Monitoring β Overview (sentiment tile), Competitors and Prompts tabs; the client portal's sentiment words; the weekly digest.
Why it matters
A sentiment trend you cannot trust is worse than none. This one now matches what a careful reader would say.
The report now shows the exact passage each cited page contributed
For every page an AI engine cited and we could read, the report now quotes the sentence of the answer that page supports and the passage on the page that supports it.
How it works
After the page checks finish, each cited page is judged twice by a decision model: first, is the page about your client or about a different organization with a similar name; second, does a passage on the page support what the AI said. Only judgments the model is confident in are shown as facts; borderline ones are marked "likely" in the agency portal and left out of the client portal entirely.
What changed
Before: You knew which pages the engines cited and whether your client was named on them.
Now: You know which sentence of the AI's answer each page backs, with the passage quoted, and which cited pages are actually about someone else.
For example
A managed-security client sees that Perplexity's claim about "24/7 monitoring for regulated firms" rests on one paragraph of a competitor's services page β the passage to answer with a page of their own.
Where to find it
Audit report β Sources ("What the AI took from it" column) and the new "Pages about someone else with your name" section; the audit page in the agency portal; the client's Monitoring β Sources tab; the client portal's Sources page.
Why it matters
A citation tells you a page mattered. The passage tells you what to write.
Fix-It artifacts and prompt suggestions now come from Claude Opus 5
The model that writes your Fix-It artifacts β schema, FAQ copy, profile drafts, title rewrites β and the one that suggests monitoring prompts moved from Claude Sonnet 4.6 to Claude Opus 5, Anthropic's strongest generally available model.
How it works
Every Fix-It artifact and every prompt suggestion is drafted by a Claude model working from the audit findings and the client's intake profile. That model is now Opus 5, which reasons through the brief before writing. Audit scoring is untouched: the five engines we query to measure visibility, including the Claude engine, are exactly as before, so scores stay comparable.
What changed
Before: Artifacts and suggestions were drafted by Claude Sonnet 4.6.
Now: They are drafted by Claude Opus 5, with more room to reason about the client's specifics before writing.
For example
A FAQPage schema for a managed-security client draws its questions from the client's actual differentiators rather than generic category questions, and the H2 rewrites read like a buyer's search, not a template.
Where to find it
Client β Recommendations (every newly generated artifact), and Monitoring β Prompts β Suggest prompts. Existing artifacts are not regenerated.
Why it matters
The artifact is the thing the client actually ships. A better writer at that step is the cheapest quality gain in the product.
The report now says which cited pages actually name your client
For every page an AI engine cited, we read it and record whether your client is named on it β and list the pages the engines trust where your client is absent.
How it works
After an audit finishes, we fetch each readable page the engines cited, once, politely and respecting the site's robots rules, then look for your client's name and its competitors' names on the page. A page where a competitor is named and your client is not is the clearest opportunity there is. Pages we cannot read β LinkedIn, Reddit, sites that block crawlers β are counted but marked as not readable, never as "not named".
What changed
Before: You knew which sites the engines cited and how often.
Now: You know which of those pages name your client, which name competitors instead, and which to go after first.
For example
A managed-security client finds that a directory's "top providers" list cited seven times names a competitor and not them. That list is the first call to make.
Where to find it
Audit report β Sources ("Names you" column) and the new "Pages the AI trusts where you're absent" section; the audit page in the agency portal; the client's Monitoring β Sources tab.
Why it matters
This is the line that turns a visibility score into a to-do: the AI trusts this page and your client is not on it.
Every audit now shows where the AI engines got their answers
A new Sources section lists the sites each engine cited, sorted by how often, and says which are your client's own site, a competitor, a directory or review site, or somewhere else.
How it works
When an engine such as Perplexity or Google AI Overviews answers a question, it cites the web pages it read first. We already recorded those citations with every audit and every monitoring run; now we sort them by domain and classify each one. Your client's own site is matched by its website address, competitors by the competitor list on the client record, and directories such as G2, Clutch and Gartner from a maintained list. Platforms we cannot read, like LinkedIn and Reddit, are still counted but flagged as such. The report shows the share of citations in each group and the most-cited domains per engine. In monitoring, the Sources tab now also lists which domains appeared or disappeared between one run and the next.
What changed
Before: The report listed the five domains Perplexity cited most and nothing else.
Now: Every engine that cited anything gets its own table, with each domain typed, plus the overall split between your client, competitors, directories and third parties.
For example
A managed-security client learns that 97% of the pages the engines relied on are not theirs, that two competitors' sites are cited more than their own, and that a Star Wars wiki is the single most-cited source β a sign the engines are confusing the brand with something else.
Where to find it
Audit report β Sources (HTML, PDF, Word and Markdown), the audit page in the agency portal, and the client's Monitoring β Sources tab.
Why it matters
A score tells a client where they stand. A source tells them where to act: the specific directory to claim, the competitor page to answer, the publisher to pitch.
ChatGPT scores were being held down by responses we never received
About half of all ChatGPT answers were arriving empty and being scored as though the AI had answered and never mentioned your client. Those are no longer counted, so ChatGPT scores will rise.
How it works
ChatGPT's newer models think before they answer, and that thinking draws on the same budget as the reply itself. Our budget was set too low, so on roughly half of all calls the model used it up reasoning and returned nothing at all. Nothing in the response said anything had gone wrong, so we recorded it as a normal answer with no mention of your client β a zero. We have raised the budget so the model can finish, and more importantly we now treat an answer we never received as missing data rather than as a score of zero. An answer that genuinely arrives and does not mention your client still scores zero, exactly as before.
What changed
Before: Roughly half of ChatGPT answers silently counted as "your client was not mentioned", pulling the ChatGPT score and the overall score down.
Now: Only answers we actually received are scored. Calls that fail are recorded as failures and left out of the average.
For example
A client whose ChatGPT answers were half empty was being scored on a mix of real answers and blanks. The blanks are gone, so the ChatGPT figure now reflects only what ChatGPT actually said.
Where to find it
Audit report β Platform-by-Platform, and the ChatGPT figure on the client dashboard.
Why it matters
ChatGPT is the engine clients ask about first. Its score was understated in every audit run before today, which made progress there look worse than it was.
Outreach targets now come with a way to reach the publisher
Each page worth pitching carries the contact route the publisher advertises β the by-line, a submissions page, or an editorial address β so you are not hunting for a human before you can start.
How it works
The report already names the pages shaping AI answers where your client is absent. Acting on that still meant opening each one and looking for someone to contact. During the audit we now read the specific page each engine cited and pull out what the publisher has deliberately published for this purpose: who wrote the piece, any "write for us" or submissions link, and role addresses such as an editorial or tips mailbox. Referencing the actual author of the actual article is a far stronger pitch than a generic contact form.
What changed
Before: You got a ranked list of pages and then went looking for the right person on each one by hand.
Now: Most targets arrive with a name or an editorial route attached, and the ones that do not say so plainly.
For example
A cited round-up now shows the by-line and the publication's editorial address underneath the domain, so the pitch can reference the author's piece rather than land in a general inbox.
Where to find it
Audit report β Where To Get Mentioned, beneath each source. Also in the Markdown and Word exports.
Why it matters
Getting mentioned on third-party pages is the strongest lever on AI visibility in the published research. Finding the human was the last manual step in it.
Reports now say what moves each engine, not just where the client is weak
Every underperforming engine carries its own short remediation list, because the engines do not draw on the same sources and the same score needs a different fix depending on which one produced it.
How it works
Published research found only 7 of the 50 most-cited domains appear across Google AI Overviews, ChatGPT and Perplexity β a 14% overlap. The practical consequence is large: Perplexity draws roughly 28.6% of its citations from pages already ranking in Google's top 10, while ChatGPT overlaps with those results only 8 to 10% of the time. So a weak Perplexity score usually means an SEO problem, and a weak ChatGPT score usually does not β it points at Reddit, YouTube and third-party roundups instead. Each engine scoring below the strong band now shows a short list of what actually moves it.
What changed
Before: The report told you which engines were weak and left you to guess whether the same remedy applied to all of them. It usually does not.
Now: Each weak engine carries its own levers, in priority order, with the reasoning behind them.
For example
A client scoring 30 on ChatGPT and 72 on Perplexity used to get one undifferentiated to-do list. Now ChatGPT carries advice about consensus surfaces and explicitly warns that improving Google rank will not fix it, while Perplexity β already strong β is left alone.
Where to find it
Audit report β Platform-by-Platform Findings, inside each engine's block. Also in the Markdown and Word exports as "What Moves Each Engine".
Why it matters
Generic AEO advice sends agencies to work on the wrong thing. Knowing that SEO carries over three times more strongly to one engine than another changes where the next month of effort goes.
Reports now open with what your client can fix before lunch
A new section lists every recommendation that can be live within the hour, ordered by how fast it ships rather than how much it matters.
How it works
The roadmap answers "what matters most". It could not answer "what can we do today", because its timings are derived from effort and never drop below one to two weeks β so a five-minute robots.txt edit read as two weeks of work. Time-to-live is now tracked separately: effort measures work, time-to-live measures how long until the change is actually published. Getting listed in a third-party roundup might be ten minutes of writing and three weeks of waiting on someone else's editorial calendar, and those are very different things to put in front of a client. Anything live within the hour is pulled into its own section at the front, fastest first.
What changed
Before: Every recommendation carried a timeline of at least one to two weeks, so the quick technical fixes were buried among things that take a month.
Now: The fixes you can ship this morning are listed first, each with a realistic time, and the roadmap still covers everything by importance.
For example
A client blocking Perplexity in robots.txt now sees "5 minutes β Restore answer-engine crawler access" at the top, above a 30-minute schema fix and a 45-minute heading rewrite. Under two hours of work, all of it live today.
Where to find it
Audit report β Do This Before Lunch, immediately before the roadmap. It appears in the HTML report, the Markdown export and the Word export.
Why it matters
The most common question after a report lands is "where do we start on Monday". Ranking by importance answers a different question, and the fast wins were getting lost.
You now get alerted when an AI engine stops answering entirely
Monitoring raises an alert when a tracked engine returns nothing usable for a whole run β so an engine going dark is a notification rather than a number that quietly drifts down.
How it works
When an engine errors, monitoring records the failure and carries on, so one broken integration can never sink an entire run. That resilience is right, but it has a side effect: an engine that fails every single call looks like a slightly lower score rather than an outage. There is now a rule for exactly that case. If a tracked engine produces no usable answer at all for a run, you get an alert naming the engine and the reason. It only fires on a total failure β one good answer means the engine is working, and a run that never dispatched anything (quota exhausted, no prompts) stays silent so you are not alerted about a day that never happened.
What changed
Before: An engine could fail every call for weeks and simply show as reduced visibility, indistinguishable from your client genuinely losing ground.
Now: The engine going dark is its own alert, with the provider error attached so you can tell a billing problem from an outage.
For example
An engine's API account runs out of credit. Instead of a slow unexplained slide in that engine's share of voice, you get: "ChatGPT returned no usable answers in 3 attempts β 429 insufficient_quota".
Where to find it
Alerts, alongside visibility drop, share-of-voice overtake and lost citation. It appears in your alert rules list and can be switched off there like any other rule.
Why it matters
A dead engine and a declining client look identical on a chart. Telling them apart is the difference between chasing a content problem that does not exist and fixing an integration.
Daily monitoring now asks each prompt three times, and your plan credits went up to match
Monitoring runs sample every prompt three times per engine so a brand appearing intermittently is visible as intermittent. Plan credit allowances were tripled at the same time, so nobody loses a day of coverage.
How it works
A credit has always meant one question asked of one engine on one day, and that has not changed β a sampled run genuinely uses three. Left alone, that would have burned through a monthly allowance in roughly a third of the time and quietly stopped monitoring partway through the month. So every plan cap was tripled in the same release. Your coverage is exactly what it was; the number next to it is simply counting real work rather than being redefined to hide it.
What changed
Before: A brand that shows up in an answer roughly half the time looked like it appeared one day and vanished the next, and there was no way to tell that apart from a real change.
Now: The same brand reads as consistently intermittent, which is a different and far more useful finding.
For example
A client hovering on the edge for their main category prompt used to produce a jagged daily chart that was mostly sampling noise. It now reflects how reliably they actually appear.
Where to find it
Monitoring dashboards and the client portal β no change to how you read them.
Why it matters
Day-to-day monitoring is where agencies spot movement, and noise that looks like movement wastes their time and their credibility with the client.
Every prompt is now asked three times, and the report says how firm the number is
Scores are averaged across three independent runs per engine instead of one, and the report shows how often your client actually appeared rather than implying the number is fixed.
How it works
Answer engines do not give the same answer twice. Ask ChatGPT the same question three times and your client may be named twice β published research found only about a third of brands stay visible from one answer to the next. Asking once and reporting the result as a measurement meant the score moved between audits for reasons that had nothing to do with the client. Each prompt now goes to each engine three times, the runs are averaged, and the spread is recorded. Every run is stored, so the evidence of that volatility is kept rather than discarded.
What changed
Before: A prompt that genuinely sat on the edge could score 58, 31 or 55 depending on which single answer we happened to catch β and the report presented whichever one it got as the fact.
Now: The same prompt reports its average with the range visible, plus how many runs out of three actually named your client on each engine.
For example
A client sits on the boundary for "best managed security provider". Their report now reads: ChatGPT named in 3 of 6 runs (50%), Perplexity 5 of 6 (83%), Gemini 0 of 6 β a far more useful conversation than a single number that will look different next month.
Where to find it
Audit report β Platform-by-Platform Findings. Each engine now shows a Consistency figure alongside its other signals, and Methodology & Confidence explains how many runs produced the scores.
Why it matters
Measurement is the product. A number that swings by twenty points between audits because of sampling noise costs an agency credibility with their client, and there was no way to tell that noise apart from real movement.
SACS moves to v2.0, and certificates you already hold stay valid
The certification standard now requires that signals be visible without JavaScript. Certificates issued under v1.0 keep their version, keep their badge, and stay valid until their original expiry date.
How it works
Because audits no longer count signals that only appear after JavaScript runs, a fully client-rendered site that certified under the old rules may not certify under the new ones. That is a change to the standard rather than a correction, so it would be wrong to apply it retroactively. Certificates have always been issued one per qualifying audit and never rewritten, so a certificate you already hold is untouched by a later audit β and each one now records which version of the standard it was awarded under. The public verification page and the badge both report that recorded version, so a v1.0 certificate is presented honestly as a v1.0 certificate.
What changed
Before: Every certificate was displayed as v1.0 because the version was hardcoded, with no way to tell which rules a given certificate was actually granted under.
Now: Each certificate carries its own version. Existing ones read v1.0 and run to their original expiry; new ones are issued under v2.0.
For example
A client certified in March on a React site keeps their badge and their verification page until it expires. Their next audit is scored under v2.0, so if the site is invisible without JavaScript they will see that in the report before the certificate lapses β with time to fix it.
Where to find it
The version appears on the SACS badge itself and on the public verification page any prospect can reach by clicking it.
Why it matters
A certification is a promise about a moment. Silently re-judging past certificates against new rules would break that promise, and quietly relabeling an old badge with a new version number would claim something that was never granted.
Audits now check what AI crawlers see with JavaScript switched off
Every page is fetched twice β once in a browser and once the way an answer engine fetches it β so the report can tell you when content or schema only exists after JavaScript runs.
How it works
Our crawler uses a real browser, which means it sees pages exactly as a visitor does, after all the JavaScript has run. Most AI crawlers do not run JavaScript at all. For a site built on a JavaScript framework those are two very different pages, and until now we were only ever looking at the friendlier one. Each crawled page is now also fetched plain, with no browser, and every on-page signal is scored from that version instead: structured data, E-E-A-T signals like case studies and author bios, outbound links to places like G2, and question-style headings. If a page cannot be fetched plainly we fall back to the browser version, so an unreachable page never costs you credit you genuinely have.
What changed
Before: A site could pass the technical section with full marks while being almost blank to ChatGPT and Perplexity, and nothing in the report would hint at it.
Now: The report says how much of each page survives without JavaScript, names the worst-affected pages, and flags any schema the engines cannot see.
For example
A client's FAQ schema is injected by their tag manager. It is genuinely on the page in a browser, so it used to count toward their score β but no answer engine ever saw it. It is now scored as missing, with the reason spelled out.
Where to find it
Audit report β Technical Gap Analysis β Crawler Access, as "JavaScript-free content". Affected schema rows read as missing and explain why.
Why it matters
This was the difference between a report describing the site your client sees and a report describing the site the answer engines see. A score built on signals the engines never receive is a comfortable number that predicts nothing.
Reports now name the exact pages to get your client mentioned on
A new "Where To Get Mentioned" section lists the sources the AI engines actually drew on for your client's category, ranked by how often they appeared in answers that never mentioned the client.
How it works
Every audit already records which sources each engine cited, for every prompt. It also already knows which answers named your client and which did not. Putting those two together gives something neither one gives alone: a domain cited ten times in answers that already mention your client is not an opportunity, but a domain cited ten times in answers where they are absent is exactly where being listed would change the result. The section ranks by that absence. Each source is tagged with the play that actually fits it β pitch an editorial listicle, claim a review profile, participate on Reddit as a named person, create or sponsor a video, correct a reference entry β because treating all five the same is how brands get themselves banned from Reddit.
What changed
Before: You knew a client was invisible for a set of prompts, but working out where to go about it meant opening each answer by hand and reading the sources one at a time.
Now: The report hands you a ranked shortlist of the pages already shaping those answers, with the sample prompts to quote in the pitch.
For example
A managed-security client is missing from answers about mid-market providers. The section shows TechRadar cited across ChatGPT, Perplexity and Gemini in three answers that never mention them β one placement sitting in front of all three engines at once.
Where to find it
Audit report β Where To Get Mentioned, after the citation analysis. It appears in the HTML report, the Markdown export and the Word export.
Why it matters
Getting mentioned on third-party pages is the single strongest lever on AI visibility in the published research β stronger than backlinks or domain authority. This turns that from advice into a specific, evidenced list of who to contact.
Audits now check whether AI crawlers are allowed to read the site
Every audit reads the site's robots.txt and reports which answer engines are permitted to retrieve it β because a blocked crawler cannot cite a site no matter how good its content is.
How it works
Each engine sends its own crawler, and robots.txt decides which ones get in. We now check every one by name and separate the two kinds. Retrieval crawlers β OAI-SearchBot, PerplexityBot, Claude-SearchBot, Googlebot, Bingbot and the user-triggered fetchers β are what make a site citable at all; if one is disallowed, that engine can never recommend the brand. Training crawlers like GPTBot, ClaudeBot and Google-Extended only affect model training, and blocking them is a deliberate, perfectly reasonable choice. We only raise a finding for the first kind, and note the second without calling it a problem.
What changed
Before: A client could score zero on an engine with no visible explanation, and the audit had no way to tell you the crawler was simply turned away at the door.
Now: The report names the blocked engines, explains what it costs, and hands you the corrected robots.txt block.
For example
A site sits behind Cloudflare, whose AI-crawler control is on by default. Its robots.txt looks open at the top, but further down it quietly disallows OAI-SearchBot and PerplexityBot. The audit now reports that ChatGPT and Perplexity cannot retrieve the site β a five-minute fix that nobody had chosen to make.
Where to find it
Audit report β Technical Gap Analysis β Crawler Access, at the top of the section. When something is blocked, a matching fix appears in the recommendations.
Why it matters
It is the cheapest win in the whole audit. Every other recommendation assumes the engine can reach the page it lands on, and this is the one finding that says whether it can.
Scan a prospect without adding them as a client
Run an AI visibility scan on any prospect website from inside the portal, without creating a client record or using your public lead-capture form.
How it works
Your public free-audit form is built for inbound: a prospect fills it in and asks for their own report, and it is rate-limited to keep it from being abused. Using it to research prospects meant fighting those limits and filling your Leads list with entries that were never leads. Prospect Scan is the other half of that job. Because you are signed in, none of the public form's limits apply β there is only a generous per-agency daily cap. Scans appear in your Leads list marked "Scan", and convert to a client in one click exactly like an inbound lead.
What changed
Before: Researching a prospect meant submitting your own public form and hitting a rate limit after the first one, or creating a throwaway client you would later have to clean up.
Now: Scan any site from the portal, as many as you need, with nothing to clean up afterwards.
For example
Before a call with a prospect, scan their site, open the report, and walk in knowing exactly how they show up in AI search.
Where to find it
Pipeline β Leads. Click "Scan a prospect", enter a brand name and website, and it appears in the list marked "Scan" once it finishes.
Why it matters
The scan-to-client path is how agencies actually win work, and it was running through a form built for a different job.
Competitors you named for audits no longer go unmonitored
The Competitors tab now tells you when rivals from a client's audit settings are missing from monitoring, and adds them in one click.
How it works
A client has two competitor lists, and they were never connected. The one on the client form drives audits β it shapes the comparison prompts your reports are built from. The one on the Competitors tab drives monitoring β mentions, share of voice, alerts. Adding a rival to one never added it to the other, and nothing said so. Now, whenever the audit list contains a competitor that monitoring is not tracking, a banner on the Competitors tab names them and offers "Add to monitoring". Matching ignores case and spacing, and counts aliases: if you already track "Pareto" with "ParetoHealth" as an alias, ParetoHealth is not reported as missing. Clicking the button creates the missing brands only, so pressing it twice does nothing the second time.
What changed
Before: You entered three rivals when you set the client up, turned monitoring on, and monitoring silently tracked none of them β the two lists looked like one list.
Now: The Competitors tab says which ones are missing and why, and one click brings them across.
For example
A client set up with ParetoHealth, Captive Resource and Benecon in its audit settings, and zero monitored competitors, now shows all three by name with a single button to start tracking them.
Where to find it
Client β Competitors tab, at the top, whenever there is something missing.
Why it matters
Share of voice is only as honest as the set of brands it compares against. A client whose real rivals were never monitored has been reading a competitive picture with the competition left out β and until now there was no way to notice.
Client portal groups fixes by where the work happens
The recommendations list in your client's portal is now split into the same three tracks the audit report uses β on your site, third-party profiles, off your site β instead of one flat list.
How it works
Every recommendation already carries a track: work your developer ships on the site (schema, copy rewrites), listings you submit to directories like G2 and Clutch, and signals you earn off-site (citations, mentions, reviews). The PDF and HTML audit report has always rendered those as three labeled sections; the client portal was showing them as one undifferentiated list. Now the portal groups them under the same three headings, in the same order, using word-for-word the same labels as the report. A heading only appears when there is something under it, so a client with only off-site work sees one section, not three with two empty.
What changed
Before: A client opened their Progress page and saw every measured fix in one flat list, with no signal that some items were their developer's job, some were a directory submission, and some depended on third parties.
Now: The same fixes are grouped under "On your site β deploy these artifacts", "Third-party profiles β submit these listings", and "Off your site β earn these signals", matching the report they already have a PDF of.
For example
A client who has shipped FAQ schema, submitted a G2 profile, and earned a press citation now sees three clearly separated sections instead of three adjacent cards that look interchangeable.
Where to find it
Client portal β Progress β "What your agency has been doing".
Why it matters
When a client reads the PDF and then opens the portal, the two now say the same thing in the same words β so the conversation on a review call is about results, not about why the two surfaces look different. It also makes the harder, slower off-site work visibly distinct from the quick on-site wins, which is the part clients most often assume should already be done.
A step-by-step guide to operating the portal
A new "Using the Portal" guide walks through the actual jobs you do here β adding a client, filling in intake, running an audit, working the fix list, turning on monitoring, and collecting free-audit leads β in plain language, with the exact click path for each.
How it works
The guide is organized by task rather than by feature, in the order you would hit them with a new client: getting oriented on the Portfolio, why your tabs may differ from a teammate's, adding a client, what each client tab is for, the intake questionnaire, running an audit, reading the report, working recommendations, the difference between Measure Impact and Mark shipped, the monitoring setup wizard, managing tracked prompts, sources and competitors, alerts, inviting a client to their portal, setting up the free audit tool, working the leads it produces, team and roles, and your white-label settings. Each section says where to find it, what it is, how to use it, and β where it matters β a callout for the thing that usually trips people up. A contents list at the top jumps straight to any section.
Where to find it
Resources β Portal guide in the sidebar (next to Report guide). On mobile it sits under the More tab.
Why it matters
New team members can get productive without shadowing someone, and you stop re-explaining the same five workflows β the ones that are genuinely non-obvious, like why intake changes what the audit asks, or why a fix shows Mark shipped instead of Measure Impact.
A plain-English guide to every part of the audit report
A new "How to Read This Report" guide explains what each section of the AI Search Visibility Audit means, in plain language, right next to the report while you review it with a client.
How it works
The guide walks through all 19 sections of the report in order β the cover page, the composite score, each of the five weighted dimensions, the AI Readiness Scorecard, platform-by-platform findings, what AI actually said about the brand, how the brand compares to competitors, who each engine cites, unprompted visibility, technical gaps, the ROI-ranked roadmap, the ready-to-deploy fix-it artifacts, and the methodology and confidence notes β with a short "what it is" and "how to read it" for each one, in plain terms with no jargon.
Where to find it
Open any completed audit β "How to read this report" (opens in a new tab so you can keep it beside the report while reviewing with a client). Also under Resources β Report guide in the sidebar.
Why it matters
Client review calls go smoother when you (or the client) can self-serve what a score or section means instead of you explaining every dimension from memory β one link does the explaining for you.
A clearer home base for your portfolio and client work
Agency admin gets a Portfolio home laid out as Clients, then Needs attention, then Recent audits β plus Pipeline for free-audit leads separate from widget setup, a calmer shell with light/dark theme, and a client portal focused on Overview and Progress.
How it works
Signed-in agency screens now share a consistent shell (sidebar on desktop, bottom tabs on mobile). Portfolio is the home, in this top-to-bottom order: (1) Clients β your full roster table (desktop) or cards (mobile), with sort-by-movement when monitoring is on and a column picker on desktop; (2) Needs attention β monitoring clients that need a look (visibility swings, alerts, or stale runs); (3) Recent audits β latest completed and in-flight work across the portfolio. Free-audit leads live under Pipeline; Free-audit setup (embed code, form copy, CRM) lives under Agency so the two jobs stay separate. The client portal keeps white-label branding and primary nav of Overview + Progress β detailed prompts and sources stay one click away as evidence, not primary tabs. Super-admin operators get the same shell pattern with Overview / Tenants / Money groups. Theme toggle (light/dark) is available in the header.
What changed
Before: Navigation mixed free-audit inbox with setup, Home was a denser dashboard without a clear reading order, and each portal felt slightly different in chrome and mobile nav.
Now: Portfolio home reads top-to-bottom as Clients β Needs attention β Recent audits, with an explicit leads pipeline, Agency-side free-audit setup, white-label client Overview/Progress, and shared light/dark shell chrome across agency and operator portals.
Where to find it
Agency Admin β Portfolio (home): Clients at the top, Needs attention next, Recent audits below. Pipeline (leads), Agency β Free-audit setup; Client portal β Overview / Progress; Super Admin β Overview / Agencies. Theme toggle in the top bar on signed-in screens.
Why it matters
You see the full client roster first, then who needs a look, then what just ran β a faster path from overview to action in a sales demo or daily ops, without changing how audits, monitoring, or tenant isolation work under the hood.
Portal login, PDF download, and monitoring setup polish
Email-only client portal sign-in, a Download PDF button on the Progress Report, unarchive for archived prompts, honest wizard credit estimates, intent chips on suggestions, and a staleness banner on Competitors.
How it works
Client portal login now asks only for an email β the agency subdomain resolves which portal they belong to, and a magic link still arrives when they have access (same uniform confirmation either way). On the portal Progress Report page, clients can download the same branded PDF your team already generates. In agency admin: archived tracked prompts regain an Unarchive action (restores them to paused, not active, so the prompt cap is not consumed); the monitoring setup wizard Step 3 shows the real daily credit math from your plan's engine set (N prompts Γ M engines); AI-suggested prompts carry an intent chip (comparison, category, and so on) that sticks when you accept them; and the Competitors tab now shows the same amber "data may be stale" banner Sources already had when a run is more than a few days old.
What changed
Before: Portal login asked clients for a raw client UUID; Progress Report PDF lived only on the agency side; archived prompts were a dead end; the wizard could only estimate credits with vague wording; suggestion intent was discarded on accept; Competitors had no staleness signal.
Now: Clients sign in with email alone, download their own Progress Report PDF, and agencies get unarchive, real credit estimates, intent chips, and a consistent staleness banner across Sources and Competitors.
Where to find it
Client portal login; Client Portal β Progress Report β Download PDF; Client page β Prompts (Unarchive on archived rows, intent chips on suggestions); Monitoring setup wizard Step 3; Competitors tab (staleness banner when runs lag).
Why it matters
Removes the last frictions between "monitoring is live" and "a real client can use it without you babysitting the UI" β less hand-holding on login and reports, clearer cost and intent signals when you set a client up.
Progress Report trends don't cross scoring methods anymore
The score trend in the Progress Report no longer draws one line across a scoring-methodology change, and the headline delta only compares audits scored the same way.
How it works
When we tightened scoring in July (see βMore trustworthy visibility scoresβ), audits from before and after that change stopped being directly comparable β but the Progress Report's headline delta and trend line kept comparing them anyway. Now the trend chart shows the earlier method in muted gray and the current method in your brand color, with a clear gap between them, and the headline delta is calculated only using audits from the current scoring method.
What changed
Before: A client whose true score barely moved could see a large, alarming drop that was really just the scoring update β with no explanation why.
Now: The headline delta reflects genuine movement only, with a plain-language note explaining the methodology change when one is in view.
For example
A client's report previously showed "-22.3 pts" driven mostly by the July scoring update; the same audit history now reads as a small, honest improvement within the current method.
Where to find it
Client Portal β AI Visibility Progress Report, and the same PDF/HTML report from the client page.
Why it matters
You never have to explain away a fabricated regression to a client β the number they see is one you can stand behind.
Your clients get their own portal
A read-only, agency-branded portal your monitoring clients log into with a magic link β Overview, Prompts, and Sources from their monitoring data, plus the AI Visibility Progress Report. Managed from the client page.
How it works
Each client gets their own portal at your agency subdomain (yourbrand.sextantaeo.com/portal), served under your branding β logo, colors, and agency name, never Sextant. Access is a magic-link email, the same allowlist-based login the platform already used, now delivered as a proper HttpOnly session cookie instead of a token in the URL. Once in, a client sees three read-only tabs built from their own monitoring data: Overview (visibility and share-of-voice trend vs. their tracked competitors), Prompts (every tracked prompt with plain-language visibility and sentiment, and the actual AI answer text on demand), and Sources (the domains AI engines cite about them, with trend). A fourth page, the AI Visibility Progress Report, shows their composite score trajectory over time, a plain-language summary per scoring dimension, and β when your team has shipped Fix-It recommendations β the Closed-Loop Proof of what was done and how visibility moved afterward. None of it is editable: no run controls, no competitor management, no cost or credit figures, no engine-failure noise β only the client's own aggregated numbers.
What changed
Before: Clients had no direct view into their own monitoring data β they saw it only through whatever the agency chose to screenshot, export, or walk them through manually.
Now: Clients log in on their own with a magic link and see a live, agency-branded read-only view of their monitoring and Progress Report, updated as new runs complete β no agency staff time spent relaying numbers.
Where to find it
Client page β new "Portal access" panel: invite a client by email, resend or revoke access, and see who has logged in and when. A "Preview as this client" button opens the same portal in a new tab exactly as that client would see it, for a quick QA check before or after an invite. The client's own login is at yourbrand.sextantaeo.com/portal.
Why it matters
This is the differentiator neither Otterly nor Peec offers today: your client-facing brand extends all the way to the people you report to, not just your own team, without you building or maintaining anything yourself.
Set up monitoring in minutes
AI-generated prompt suggestions from the client profile, competitor auto-discovery from the client's actual AI answers, CSV bulk import, and a guided setup wizard that walks a new monitoring client through brands, prompts, and the first run.
How it works
On the Prompts tab, a Suggested prompts panel generates a short list of AI-visibility prompts from the client's profile β name, website, industry, services, and intake answers β for one-click Accept or Dismiss; nothing is added to the tracked list until you accept it. On the Competitors tab, a Find competitors button reads the client's own recent AI monitoring responses and surfaces brand names that show up repeatedly but are not yet tracked, again as Accept/Dismiss suggestions. The Prompts tab also gets a CSV import: pick a file, preview the parsed rows, import, and see a per-row results summary of what was added and what was skipped and why. For a brand new monitoring client with nothing tracked yet, the Monitoring tab now opens a three-step guided wizard instead of an empty dashboard: confirm your brand and competitors (prefilled from the client record), review and accept AI-suggested prompts or add your own, then launch the first monitoring pass.
What changed
Before: Setting up monitoring for a new client meant manually typing every prompt and every competitor into the Prompts and Competitors tabs one at a time, with no guidance on what to track first.
Now: A new monitoring client sees a guided wizard that prefills brands from the client record, generates a starter set of prompts from the client profile, and launches the first run β and any client can add AI-suggested prompts, auto-discovered competitors, or a CSV of prompts at any time after that.
Where to find it
Client page β Prompts tab (Suggested prompts panel, CSV import) and Competitors tab (Find competitors). The setup wizard appears automatically on the Monitoring tab for a monitoring-enabled client with nothing tracked yet.
Why it matters
The gap this closes is time-to-first-value: an agency can go from "monitoring enabled" to "first real data" in minutes instead of a manual setup session, and the ongoing suggestion/discovery tools mean prompt and competitor lists stay current without agency staff having to think of everything themselves.
Continuous AI visibility monitoring
Track how a brand shows up in AI answers every day, not just at audit time β across ChatGPT, Claude, Gemini, Perplexity, Grok, and Google AI Overviews, with trend charts, source intelligence, competitor benchmarking, and alerts.
How it works
For clients on a monitoring plan, a standing list of tracked prompts runs daily against all six AI surfaces and rolls up into four new client-page tabs. Monitoring shows visibility, share of voice, average position, and sentiment trend charts for the brand and its named competitors. Prompts lists every tracked prompt with its daily results and drills into the actual AI answer text for any run. Sources tracks which domains AI engines cite, with new/lost/rising/falling movers and a gap-analysis callout for cited sources that are not yours. Competitors benchmarks the brand against each named competitor on the same four metrics. Threshold alerts (a visibility drop, a competitor overtaking on share of voice, a citation lost) fire automatically and go out by email, alongside a weekly digest summarizing the week for every monitoring-enabled client. The agency dashboard gets a portfolio view of every monitoring client sortable by movement, plus a credit quota meter so you can see usage against the monthly cap before you hit it.
Where to find it
Any client with monitoring enabled shows four new tabs on their client page β Monitoring, Prompts, Sources, Competitors β ahead of Audits & Score. The agency dashboard shows the portfolio movement view; a credit-usage chip appears in the header once you pass 80% of your monthly cap, and the full quota meter lives in Settings.
Why it matters
Audits are a snapshot; monitoring is the trend line β the difference between telling a client "here is where you stood on the day we checked" and "here is what changed this week, and why." It is also the feature set that puts this platform head-to-head with dedicated AI-monitoring tools like Otterly and Peec, inside the same platform you already use for audits and Fix-It recommendations.
Free-audit reports now break the score down dimension by dimension
The {{score_breakdown}} report section now renders a complete card per dimension: what it measures, the score out of its maximum (e.g. 19.1/20), and a plain-English sentence explaining why.
How it works
Any report template that includes the {{score_breakdown}} token renders the five AI-visibility dimensions β Brand Recognition, Market Position, Presence Quality, Brand Sentiment, and AI Share of Voice β each as one card: the dimension name with its point weight, the question it answers (e.g. "Does an AI mention your brand unpromptedβ¦?"), the score out of its maximum, and a one-sentence explanation matched to how the brand actually performed. The default report template already includes it; add the token to a custom template to get the same section.
What changed
Before: Reports using the breakdown showed bare numbers with no maximums, in arbitrary order, so an 8.4 could read as weak when it was actually 8.4 out of 10 β and dimension definitions lived in a separate section, if at all.
Now: Each dimension is a single card: "Brand Recognition (20 pts)", what it measures, "Brand Recognition Score = 19.1/20", and a sentence explaining what that score means for the brand β without exposing scoring internals.
Where to find it
Free Audit β Settings β report template: include the {{score_breakdown}} token. Visible in prospect reports and the View Report page.
Why it matters
Prospects (and your sales team) instantly see where a brand is strong and where the gaps are β each gap sentence is a ready-made talking point for the services you sell.
Open any free-audit lead's report in one click
Every scored lead on the Free Audit page now has a View Report link that opens the full audit report the prospect received.
How it works
On the Free Audit page, the Leads table has a new Report column. For any lead whose audit has finished scoring, click View Report to open their full report in a new tab. The same link appears in the lead detail panel alongside Resend email and Convert to client.
What changed
Before: You could see a lead's score and resend their email, but there was no way to open the actual report they received without digging it out of the email.
Now: One click on View Report opens the lead's full audit report in a new tab, straight from the Leads list or the lead detail panel.
Where to find it
Agency Admin β Free Audit β Leads β View Report.
Why it matters
Before a sales call, you can read exactly what the prospect saw β their scores, their gaps, the recommendations β so your pitch picks up right where their report left off.
Recommendations now have their own tab
Every client page now has a dedicated Recommendations tab with the full generated fix on each card, one-click copy, and an honest status for Measure Impact.
How it works
The client page is now split into Overview, Recommendations, and Intake tabs. Each recommendation card shows the complete generated fix β not just a summary β with a copy button for dropping it straight into a CMS or ticket. The Measure Impact button now reflects the real state of the job: measuring, not detected live yet, or on cooldown, instead of spinning indefinitely.
What changed
Before: Recommendations were buried in the client overview, showed truncated fix text, and Measure Impact just spun with no indication of what was actually happening.
Now: Recommendations have their own tab with the full fix text, a one-click copy button, and a status-aware Measure Impact action that tells you exactly what state the job is in.
Where to find it
Agency Admin β Clients β open a client β Recommendations tab.
Why it matters
Less digging to find and act on a fix, and no more guessing whether Measure Impact is stuck or working β the status is always honest about what state the job is in.
Prove your fixes worked
Deploy a Fix-It recommendation, click Measure Impact, and see the verified beforeβafter movement per engine in the AI Visibility Progress Report.
How it works
Once a recommendation is shipped, a 'Measure Impact' button appears next to it. Clicking it re-checks the live site, then re-runs the audit and compares the targeted dimension's score on each AI engine to the baseline. The result β a composite delta, per-engine movement, and how many days it's been live β shows up as a Proof of Impact card on the client and in the Progress Report.
What changed
Before: You could tell a client you shipped a fix, but had no evidence it actually moved the needle with any AI engine.
Now: Every measured fix gets its own Proof of Impact card: composite delta, per-engine movement on the dimensions it targeted, and verified-live timing.
For example
After shipping FAQ schema for a client, Measure Impact shows 'Presence Quality moved +4.0 on Claude, verified live 12 days ago' β a concrete result to bring to the next check-in.
Where to find it
Agency Admin β Clients β open a client β Recommendations. The client portal shows the same Proof of Impact card, read-only.
Why it matters
This closes the loop: recommendations stop being a to-do list and start being evidence. It's the difference between 'we fixed this' and 'we fixed this, and here's what changed.'
Dashboard now nudges you to complete client intake
Any client whose intake form hasn't been filled out now shows a clear 'Complete Intake' prompt on the agency dashboard, linking straight to the form.
How it works
On the agency dashboard, each client card that is missing an intake profile shows an amber 'Complete Intake' button. Clicking it takes you directly to the Intake section of that client's page β no hunting through menus.
Where to find it
Agency Admin β Dashboard β look for the amber prompt on any client card that lacks an intake profile.
Why it matters
A completed intake is the single biggest lever for audit accuracy: it grounds every AI-visibility question in real facts about your client β their real competitors, real services, real differentiators β instead of guesses. The nudge makes sure nothing slips through with a hollow profile.
More secure session handling
Login sessions are now stored in a way that is inaccessible to browser scripts β an industry-standard upgrade that removes an entire class of credential theft risk.
How it works
Your login token is now held in an HttpOnly cookie, which the browser sends automatically on every request but will not expose to JavaScript β not even to code running on the same page. The platform also validates the origin of every state-changing request, so a malicious third-party page cannot trigger actions on your behalf. Nothing about how you use the platform changes.
Why it matters
If a rogue script ever ran on any page of the platform, it previously could have copied your login token. After this change that is no longer possible β your session is protected even if something else on the page misbehaves. It's the same approach used by modern banking and enterprise SaaS products.
See what weβre building next
A new Roadmap page shows what weβre working on now, whatβs next, and what weβre exploring β so you always know where the platform is headed.
How it works
Open the Roadmap from the top navigation. Items are grouped into Now, Next, and Later, each with a short description and why it matters for you. A dot on the Roadmap nav link appears when the plan changes, and clears once youβve looked.
Where to find it
The βRoadmapβ link in the top navigation, next to Whatβs New.
Why it matters
You can see whatβs coming and plan your client work around it β and when something on the roadmap ships, it moves over here to Whatβs New.
More of your configured questions run in every audit
The audit now prioritizes the use cases, problems, and tasks you enter for a client, so more of them actually get asked across the AI engines.
How it works
Every audit runs a fixed number of questions. We now reserve a dedicated share of that budget for the questions you set in a client's AI question targets β they're asked first, and competitor comparisons flex to fit around them.
What changed
Before: Competitor-comparison questions could crowd out the specific questions you had configured, so only a few of them ran.
Now: Your configured questions get a guaranteed share of each audit β up to 8 β with competitor comparisons filling whatever is left.
For example
A managed-security client with several use cases and problems entered now sees questions like 'Recommended tools for ransomware exposure' and 'Best tools for SOC 2 readiness' actually run, instead of being dropped.
Where to find it
Automatic β fill in a client's Intake β AI question targets and the next audit prioritizes them.
Why it matters
The questions you take the time to configure are the ones that matter most for your client β now they reliably get asked, and it never changes the comparable /100 score.
Ask AI the questions your client's buyers actually ask
Turn your client's real use cases, problems, and tasks into targeted questions we put to every AI engine β not just their brand name.
How it works
In a client's Intake section there's a new 'AI question targets' group. Enter the use cases, problems, and tasks their buyers care about (and, for software clients, known issues and error messages). Each becomes its own question across all five AI engines β up to three per field, so cost stays predictable.
What changed
Before: Audits asked about the brand, its industry, and its competitors β but not the specific problems and use cases buyers actually search for.
Now: Audits also probe the real buyer questions you define, so you see where your client shows up in the moments that drive new business.
For example
For a managed-security client, entering 'ransomware exposure' and 'SOC-2 readiness' adds questions like 'Recommended tools for ransomware exposure' β revealing whether the AI engines name your client when a buyer asks.
Where to find it
Agency Admin β Clients β open a client β Intake β the "AI question targets" group.
Why it matters
These are the questions that actually surface vendors to buyers. Filling them in shows your client exactly where they stand on the searches that matter β and it never changes the comparable /100 score, so your trend lines stay honest.
Introducing Whatβs New
This page β a running log of what we ship on the platform and how each change helps you.
How it works
We post here whenever we improve the platform: new features, accuracy improvements, and fixes, each written to explain what changed and why it matters for you. A dot on the 'What's New' nav item appears when there's something you haven't seen yet, and clears once you read it.
Where to find it
The 'What's New' link in the top navigation.
Why it matters
You'll always know what's changed and why it matters for the reports you deliver β no more silent updates.
Client intake profiles
Tell us the real facts about a client once, and every audit and AI-generated fix is built on the truth instead of guesses.
How it works
Each client now has an Intake section with four groups β Company (aliases, founded year, team size, client count, service areas), Positioning (category, one-liner, ideal customer, services, differentiators, certifications), Competitors (their real rivals, by name and domain), and Proof (review profiles, notable clients, awards). Fill in as much or as little as you know; it saves per client and feeds straight into the audit.
What changed
Before: The audit only knew a client's name, industry, and website, so the AI writeup filled the gaps with plausible-sounding guesses.
Now: The audit is grounded in the facts you enter β the accuracy guard lets real details through while still stripping invented ones.
For example
For a two-person managed-security shop, the report used to imply a large enterprise β 'serving 400β4,000-employee organizations.' With intake filled in, it reflects the real size and the real regional competitors you listed.
Where to find it
Agency Admin β Clients β open a client β the Intake section on the client page.
Why it matters
The report stops inventing details about your client, and competitor comparisons use the real local players you entered β so what you hand your client is accurate and credible.
Reports never invent facts about your client
Generated fixes and copy are now checked against verified facts, so the report won't claim something about your client that isn't true.
How it works
When the AI drafts recommendations or rewrites, an accuracy guard compares every firmographic claim β client counts, team size, revenue bands, awards, logos β against the facts we actually know from the site crawl and the client's intake profile. Anything it can't verify is removed before you ever see it.
What changed
Before: The generator could assert impressive-sounding specifics it had no basis for.
Now: Unverifiable specifics are dropped; only claims backed by real data survive.
For example
A drafted line like 'trusted by over 500 organizations' is removed unless 500 is actually in the client's verified facts.
Where to find it
Visible in each audit's Fix-It recommendations.
Why it matters
You can share the AI's suggestions with confidence β no awkward moment where a client says 'we don't have 500 customers.'
Unprompted Visibility
See whether the AI engines name your client on their own β before anyone mentions the brand β the truest measure of AI discoverability.
How it works
Alongside the brand-name questions, every audit now asks open, category-style questions β like 'who are the best managed security providers in New Jersey?' β where the brand is never mentioned. We measure how often each engine names your client unprompted and headline it in the report.
What changed
Before: The score only reflected how AI described the brand once you named it.
Now: You also see whether AI surfaces the brand with no prompting at all.
For example
In a recent audit, national names appeared for more than half of the open questions while the local brand appeared in none β a clear, honest picture of the gap to close.
Why it matters
This is what actually drives new business: being named when a buyer asks a generic question. It shows clients exactly where they stand and why the work matters.
See exactly who the AI engines cite
Reports now list the specific competitors and sources each engine named, so you can show clients who's winning the category.
How it works
For every question, we capture which brands and which sources β like G2, Clutch, or a competitor's own site β each engine actually referenced, and surface the named players in the report.
What changed
Before: You saw a share-of-voice number but not who was taking the share.
Now: You see the named competitors and sources behind the number.
For example
A cybersecurity audit surfaced the exact rivals the engines kept naming β the shortlist your client is being measured against.
Why it matters
Concrete names turn an abstract score into a conversation: here's who's ahead, here's where they're cited, here's how we close the gap.
More trustworthy visibility scores
When an AI engine admits it doesn't recognize a brand, that no longer counts as visibility.
How it works
Replies that disclaim knowledge β 'I'm not familiar with that company', 'no publicly available information' β now zero out the visibility dimensions they affect, instead of scraping points from stray brand-name matches.
What changed
Before: A reply that literally said it didn't know the brand could still score around 57/100.
Now: Those brand-unknown replies score honestly β the visibility dimensions drop to zero.
For example
A client the engines didn't recognize moved from a flattering-but-false high score to one that reflects reality.
Why it matters
A strong number is now one you can defend in front of a client, and a weak one accurately shows the work to be done.
We audit your client's real pages
The crawler now discovers a site's actual pages before auditing, instead of guessing at URLs.
How it works
We read the sitemap and follow the site's own navigation to find real, live pages, checking each one resolves before including it β no more probing for pages that might not exist.
What changed
Before: The crawler guessed a handful of common paths and sometimes counted 404s as if they were pages.
Now: It finds and verifies the real pages, so recommendations are specific to what's actually on the site.
For example
On one site this went from six guessed URLs β several of them dead β to eleven real, verified pages.
Why it matters
Recommendations point at pages that genuinely exist, so the fixes you pass along are actionable rather than generic.