How can AEOTrack help you?

October 27, 2026

How the score works

What the score is

The AI visibility score is a 0 to 100 index of how your brand appears in AI answers to the questions you track. It is not a percentage of anything. A 0 means no answer in your latest check named you or listed your own site as a source; a 100 would mean every engine named you first, cited your site as a source and described you well, on every question. The Questions score (buying questions) and the Topics score (topic phrases) are computed the same way over the two different sets.

The definitions below are scoring version 10 (27 October 2026). When the formula changes, the version number changes, every stored score keeps the version it was computed under, and the trend chart marks the step. Version 10 gives an answer that lists your own site as a source its citation part even when it does not name you, and counts it in your citation rate. Version 9 (October 2026) re-read the stored answers once with three corrections: questions that contain your brand are scored on their own, apart from the headline; your site's own name counts as your brand; and a name an answer ties to another company's website is not counted. A step on the chart at a version mark is that rescoring, not a change in how often you are named. The full formula, with a worked example and the version history, is on the public methodology page. On the Dashboard, How this score works, under the tiles, opens the short version. The first tile leads with how many of your latest answers name you, with the score as its second number, and the line under the two numbers says what the score is built from: its ± band, how many answers and engines, and the date of the check. When the score covers fewer answers than the count above it, because it leaves out advice-only questions and questions with your name in them, the line says so: "45 answers in the score". The ? beside the tile's label gives the rates behind the number and how many advice-only questions were left out.


1. One live answer per engine per question

Each check asks every engine each of your questions once, word for word, with the engine's live web search on, and stores the answer whole. Nothing is asked several times and averaged. On a paid plan scheduled checks run weekly, with your pinned questions daily; on the trial every question is checked daily while you use it. A manual re-check uses your plan's monthly allowance of on-demand re-checks; one re-check of a question within 24 hours of its last check is free. These answers come from fresh API calls, so your own chat can answer differently: see Why results differ from your own chat.

2. Five engines

ChatGPT, Gemini, Claude, Perplexity and Grok, on every plan and on the trial. Each engine carries a weight that stands in for its share of the people asking; ChatGPT counts most. An engine that is not configured on the platform is not measured: it is absent from the score, never treated as a 0.

3. What one answer scores

An answer that names your brand scores 25 plus up to 75 for how you were named:

  • Prominence (45%): how high in the answer's list you appear.
  • Citation (30%): whether the engine used a page on your own site as a source.
  • Sentiment (25%): the tone of the passage that names you, rated by a model.

An answer that does not name you but lists a page on your own site as a source scores 22.5: the citation part alone, with no presence points, rank or tone, since there is no passage naming you. That is below the 25 any named answer starts from, so being named always counts for more than being only read. It also counts in your citation rate; it never counts as being named. An answer that neither names you nor lists your site scores 0.

A link to your site with no mention of your name in the text is not a mention; it counts towards the citation part only.

What counts as your name. Your domain, your brand name and your brand aliases. When you add a website, the names your own homepage gives the business (its title and its structured data) are added to the aliases when they fit the domain, so "Acme Coffee Roasters" counts for acmecoffee.example without you typing it. You can remove any alias; one you remove is not added back. Names written apart count too: "Acme Coffee" for acmecoffee.example. When the answers write your business under a name that does not count yet (one they link to your site, a name on your homepage that does not fit the domain, or a near spelling), the Dashboard asks Is this you? and adds it in one click: see Brand names and "Is this you?".

A namesake is not you. When the only name that matched is one the answer ties to another company's website (it cites, links or writes that company's address, and never yours), the answer is not counted as naming you. A more specific name of yours in the same answer still counts. If many of your answers are about a different company with your name, confirm your business so the questions are written for the right one; the Dashboard asks Not you? when at least two answers tie your name to the same other company (Answers about a different business with your name).

4. From answers to the number

Per question, the engine-weighted mean of the latest answers. Per website, a recency-weighted mean of the question scores: a question's weight halves every 30 days and reaches zero at 90, so the number is your visibility lately. The Visibility Tracker's Score column is the per-question step; the Dashboard shows the per-website step.

5. Advice-only questions are left out

A question where no engine names any business ("how do I write a privacy policy?") could never move the number, so it is left out of the headline score, the mention rate and the ± range. The count of such questions is in the ? beside the Dashboard's first tile, the row on the Questions page reads Advice-only · not in your score with a Replace action, the answers stay readable, and the question comes back into the score the moment any engine names any business for it. A question where the engines name competitors and not you is not advice-only; it is a real score (0, unless an answer lists your site as a source), and it stays in. A question that can never name you over several checks (a how-to question, or software answers when you sell a service) is flagged as This question doesn't fit, with a replacement: see A question that doesn't fit.

6. Questions that contain your brand are scored on their own

A branded question names you ("is Acme Coffee Roasters any good"), and a near-brand question contains your name once the spaces are removed. Any engine names you when it is asked about you by name, so these questions say nothing about whether a buyer who does not know you yet would find you. Since version 9 they are left out of the headline and scored on their own, as the figure labelled Branded: on the Dashboard under More detail, in The other scores, and on the Questions page beside the visibility score. In your first 14 days the Dashboard's first screen also shows it under Questions with your name in them. The headline is your score on the questions a buyer asks before knowing your name.

If every question you track is branded or near-brand, they all count, so you still have a number, and the line under the headline says so, as does the first screen under Questions with your name in them. Add a question a buyer would ask before they knew you existed. The badges are explained in Branded and near-brand questions.

7. A failed answer is never "not named"

An engine call can fail: an error, a time-out, a rate limit. A failed answer is never counted as "not named", in the score, in the "N of M answers" counts or in any email.

  • When an engine's newest answer to a question failed, its last good answer stands in. The question shows that answer with its own date and says the newest attempt failed (for example "HTTP 429" or "timed out"). The score uses it while it is less than 90 days old.
  • An engine that has never answered a question shows the error and counts nowhere.
  • An answer cut off at the length cap before it named anyone is shown but not counted, neither as a miss nor as a mention.

8. What the ± means

Every score carries a ± range with its sample count ("62 ±8, from 25 AI responses"). The ± is the range to expect between two checks of the same questions: a move inside it is normal variation, not a change in your visibility. Alerts and the "What changed" card only report moves that clear it. The details are in What the ± on a score means.

9. States that are not a 0

A blank is a state, not a 0, and the app names which one:

StateMeaning
Not enough data yetFewer than three usable answers. The first check clears this: your first answers arrive about 7 minutes after you add your website.
Advice-onlyEvery question in the set is advice-only, so nothing is left to score. Replace one with a question buyers ask by name.
StaleNo usable answer in the last 30 days. The last score stays readable, marked as of its date; a check refreshes it.
Not measuredAn engine not configured on the platform, or a question that is paused. Absent from the score, the mention rate and the range.
Newest attempt failedThe engine's newest answer failed, so its last good answer is shown with its own date (section 7).
InconclusiveAn answer cut off at the length cap, or an engine that has never answered the question. Excluded, never counted as a miss.

A 0 is reserved for an engine that answered and did not name you. If that is your first result, read Why AI does not name you yet.

10. What the free audit's grade is

The free audit's 0 to 100 grade is labelled Site hygiene: a check of the homepage's code and content, how easy the page is for AI engines to read. It is not this score and not a measure of AI visibility. The audit's AI check is one question on two engines, shown in full; the score above is every tracked question on five.