Docs › Screens

AI provider visibility

What ChatGPT, Claude, Gemini and Perplexity say about your brand — your score, who they name instead, and what to do about it this week.

#In 10 seconds

• EOMA asks four AI engines a fixed set of questions about your market every week and keeps every answer.
• The page reads top to bottom in two halves: everything above the orange line is measurement, everything below it is what to do.
• The one number to watch is the Overall AEO score, and the two figures under it matter more than the score itself.

#Before you start

• Your first collection must have finished. Until then the page says "Preparing your insights" and nothing else. Nothing is broken; the engines have not answered yet.
• Your website address must be saved in your brand profile — it is how EOMA tells "AI cited you" from "AI cited a rival". See Brand profile.
• Nothing else. This page is not restricted by plan.

#Steps

#1 — Open it from the sidebar

┌─ sidebar ───────────────────────────┐
│  Dashboard                          │
│  ▼ AI Visibility                    │
│      ● AI Provider Visibility ← here│
│        Your Brand on Google         │
│  ▶ SEO                              │
└─────────────────────────────────────┘

The parent row AI Visibility lands here too, so either click works.

#2 — Read the score, then the line under it

The score alone tells you little. The sentence under it — "knows you 84.2 · suggests you 0.0" — is the finding.

#3 — Scroll to the orange line

Above it is what was measured. Below it is what EOMA thinks you should do about it. With two minutes, read the score at the top and the three to-dos at the bottom.

#What you'll see

#The chapter line — "AI Visibility · Asking AI · ChatGPT · Claude · Gemini · Perplexity · Google AI Overview"

An indigo stripe listing exactly who was asked, then a short explainer card you can close with the ×.

#"Overall AEO score" and "Responses analyzed"

The two headline figures, top left, each with a small ⓘ for its definition.

#The switch — "Providers" / "Breakdown"

Top right. It swaps what the narrow right-hand column shows, and nothing else.

#"AEO score over time", and the narrow column beside it

The wide chart is one line per AI engine, week by week. The column beside it is either the four engines with a percentage each (Providers) or your score split across the eight kinds of question EOMA asks (Breakdown).

#"Share of Voice — Where the conversation lands"

A donut: your slice highlighted, rivals with their own, small ones folded into "More competitors". The chips above it — Self, Competitors, Others, My tracked list — change what is counted, and the ⓘ in its heading flips the card over to "How high the bar actually is".

#"Where AI looks for answers"

Beside the donut, the websites AI reads when it answers questions about your market, ranked. A different question from the donut: that one ranks who gets named, this ranks whose link the engine reaches for.

#The question list

Full width, one row per question asked, showing the engines, the question, its category, your top rival on it, and how you did. Clicking a donut slice rewrites it to that rival's rows.

#The orange line — "What to do about it · Signals & Insights"

The divider between measurement and advice. Under it, two cards: Signals ("What the AI answers are telling you this week"), drawn straight from the answers one finding at a time, and Insights ("Three things to do this week").

The Insights card is not being written at the moment. The weekly write-up behind it has not run since May 2026, so this card stays on its waiting message no matter how long you leave it. Nothing about your brand causes this and there is nothing to fix on your side. Your actual to-dos are on the Actions board, which is produced separately and is unaffected. Signals beside it is live.

A band at the very bottom links to your website audit — see Web presence.

#Reading the numbers

Overall AEO score — for example 25.3. How visible you are across AI answers, from 0 to 100. It carries no % because it is not one: the larger half of it is scored against what the strongest brand in your market actually reaches, and measured leaders top out between 10 and 47. So 25.3 is not "a quarter of the way to perfect" — it is a position relative to real rivals. The method is in How visibility is scored.

"knows you 84.2 · suggests you 0.0" — the line under the score. Worth more than the score above it. The first figure is how well AI answers when the question already contains your name; the second is how often AI brings you up when it does not. A brand at 84.2 / 0.0 is famous and never recommended — everyone who already knows you gets a good answer, and nobody discovers you. Closing that gap is what the actions at the bottom of the page are for.

Responses analyzed — for example 525, "8 runs · 28 days · 1,656 all-time". The score was computed from 525 AI answers collected over eight sweeps in the last 28 days; "all-time" is every answer ever collected for you. The window never stretches to find data, which is why a quiet month reads as "not measured" rather than as an old number dressed as today's.

A percentage per engine in the Providers column — for example ChatGPT 31.4%. Across every sweep ever run, ChatGPT mentioned your brand in 31.4% of its answers. This is a lifetime mention rate, not the score, so it is not expected to match the top of the page and a gap is not an error. Engines disagreeing with each other by a wide margin is itself a finding: the facts about you exist somewhere but are not reaching every model.

Your slice of the Share of Voice donut — for example 3.1%. Of every brand name AI dropped into these answers, 3.1% of the mentions were yours. Flip the card with the ⓘ for the other half of that sentence: if the leader in your market reaches 58%, 3.1% is a long climb; if the leader reaches 9%, your market is wide open and 3.1% is better than it looks.

"#68 of 591" in Where AI looks for answers. Of the 591 websites AI read to answer market questions, yours ranks 68th by how often it was used. Links to you from questions that already said your name are excluded on purpose — being cited when the question names you is handed to you, being cited when it does not is earned, and only the earned kind moves this ranking.

Outcome in the question list. Each row is judged in the unit its own category measures, which is why the wording changes line to line: "4 affirmed" (answers that named you and backed it up), "1 unknown" (the model named you only to say it had no information about you — not a win), "#2 of 12" (where you placed on a shortlist and how long that list was), "cited 4×" (your site was the source, which for buying-guide questions is the result). Its Score is not sentiment: you can be described warmly and still lose the shortlist.

#When it goes wrong

#Related

• How visibility is scored — what the eight question types are and how each is scored
• Share of Voice — the same competitive question asked of Google rather than AI
• Actions — where the three to-dos at the bottom become work EOMA can do