When a Buyer Asks an AI, Who Gets Named?
Buyers increasingly ask assistants instead of searching — and an assistant answers with a handful of names. Either yours is among them or it is not, and until it is measured, nobody actually knows. This page explains what AI answer visibility is, how it can be measured honestly, what a measurement contains, and what it cannot promise.
Answers Are Replacing Result Lists
Three measured facts, each from a named public source, describe the scale of the change. Quote them freely with their sources.
A result list shows twenty links and lets the buyer choose. An answer names three or four businesses and the rest are simply not in the conversation. That is a different kind of visibility, and it deserves its own measurement — separate from classic ranking, which our click-curve research covers in depth.
How an Assistant Chooses Names
When a buyer asks “who is the best roofer in Phoenix?”, the assistant does some combination of three things: it draws on what its model already holds about the world, it runs a live web search and reads what comes back, and it grounds names against the entities it can identify with confidence. The names it offers are the ones that survive all three — present in its evidence, readable in its retrieval, and unambiguous as entities.
Being findable to a model is therefore not the same thing as ranking. A site can hold a strong classic position and still be invisible to the answer layer — unreadable to retrieval, inconsistent as an entity, or simply absent from the sources the engines actually cite. Each of those is a measurable condition, which is the whole premise of measuring answer visibility rather than guessing at it.
What AI Answer Visibility Is
Four defined, measurable terms — everything else on this page is built from them.
Named
Your business appears in an answer to a buyer question. The atomic unit: every other measure aggregates it.
Prominence
Where you appear when you do: the first name offered, the second, the third, later in a long list — or mentioned in passing rather than offered as an answer. Position is measured, never averaged away.
Share of answer voice
Of all the business names a market's answers mention, the fraction that are you. A market-level reading: who owns this market's answers.
Cited as a source
Your own site appears among the sources an answer cites. You can be cited without being named — your content feeding answers that recommend someone else — and named without being cited. The two splits point at different conditions, so both are measured.
How We Measure It
Real buyer questions, per market. A battery of around twenty questions is built for one (category, city) market — direct recommendation asks, shortlists, problem-first phrasings drawn from real search terms with measured volumes, and qualifier questions. The battery is stored, versioned, and owned per market; every measurement records which battery version asked its questions.
Four engines, by name, via their official APIs. ChatGPT, Claude, Gemini, and Perplexity — asked as a consumer would ask, with live web search on where the engine offers it. Gemini's lane runs without live search, and is labeled exactly that way: “model visibility · no live search.” Google's in-search AI surfaces are a separate, differently-shaped unit and are reported separately when covered — never silently folded into the same score.
Every answer kept verbatim. Each question × engine read produces a permanent receipt: the exact question, the exact answer, the sources it cited, and the date. The receipts are the evidence the scores are computed from — and they render in the report, beside the numbers they justify.
Scored as frequencies, in concept. Being named earns more than being mentioned; being named first earns more than being named third; engines count in proportion to real-world usage; a refused or failed read is excluded from its denominator, never counted as a zero. The result is a 0–100 visibility reading with its coverage stated — how many questions, how many engines, measured when. The exact coefficients stay server-side, like all of our scoring — the methodology explains every concept at this same altitude.
See What an Assistant Says About You — Right Now, Free
You do not need our report to take the first look. Build the question below, worded the way our own battery would ask it, then paste it into any assistant you already use — ChatGPT, Claude, Gemini, Perplexity, whatever is already open in another tab. Nothing here calls an engine on our end; this is you, checking directly.
Paste it into your assistant of choice, read the answer, then say honestly what you saw:
That is one real observation — worth noting, not resting on. Published measurements find the same prompt rarely returns the same list of businesses twice, so a single good answer is a point-in-time read, not a stable outcome. The report below runs this same question, and roughly fifteen others phrased differently, across four engines, and scores the pattern across all of them — not one roll.
An answer names a handful of businesses and the rest are simply not in the conversation — that is exactly the shift this page opened with. One absence does not prove a pattern by itself either, but it is worth taking seriously if it is what you just saw. The report below checks properly: the full battery, every major engine, dated and scored, so an absence is either confirmed as a real pattern or shown to be one narrow miss.
Some phrasings get a narrative answer instead of a named list — that is itself a real, measured outcome, not a failed try. We publish our own zeros the same way: a zero is a measurement. The report below runs a whole battery of phrasings specifically to separate list-shaped questions from narrative ones, so the reading reflects the question types that actually surface names in your market.
What the Report Contains
The measurement renders as a set of instruments, each carrying its measured date and coverage.
The visibility dial
The 0–100 AI Visibility reading with its coverage line, and one bar per engine with that engine's own question count.
The answer ledger
Every question, every engine, the verbatim answer one click behind its outcome — your business highlighted where it appears, every detected business marked, sources listed.
Voice, sources, and the gap
Who owns the market's answers, which domains feed them, whether yours is among them — and the measured readiness signals separating you from the businesses the engines name, stated as an association in the market, never a promise.
The record & the trend
What each engine says when asked about you directly, kept verbatim for your verification — and the visibility trend across measurements, with the measured repeat-read spread drawn only once it has actually been measured.
Subject: rankassurance.com — our own site, measured with the same instrument members receive
Measured: November 02, 2026 · 4 of 4 engines · 8 scored questions
AI Visibility (AIV): 0 / 100
Businesses named in our market's answers: 12 — and at this measurement, we were not among them. A zero is a measurement, and we publish our own.
This is live data from our weekly self-measurement, re-read on a fixed cadence with full verbatim receipts kept. We publish it as measured because an instrument you would trust on your own business should be willing to show its own reading first.
See a real report, annotated: a real AI Answers measurement of a real market — the subject masked, every competitor name and every number as measured, with each element explained in plain language.
See a complete report, real markets: the AI Answers examples in our sample library — masked subjects, real competitors, real numbers, every element explained. The sample-report walkthrough shows what each tab holds before you open one.
What This Measurement Cannot Do
Answers vary run to run. Answer engines are probabilistic: published measurements across thousands of repeated runs find that the same prompt rarely returns the same list of businesses twice (in one 2,961-run collection, the odds were under one in a hundred). That is why scores here are frequencies across a whole question battery — the statistically defensible unit — and why a single answer is treated as a point-in-time observation, never a stable “AI rank.”
An API read is an instrument, not a screenshot. Reading the engines through their official APIs is a clean-room measurement of the default answer layer. A particular person's app session — with their history, memory, and model routing — can differ. We state this on every surface rather than burying it.
The cadence is monthly. Each market re-measures on a monthly rhythm; intra-month swings between measurements are not captured, and the report says so.
And the sentence that governs everything: we measure who is named. No one can promise a place in an AI's answer, and we never do.
The Measurement Field, Documented
For completeness, the field of AI-visibility measurement tools as it stands — names, coverage, and published entry prices, recorded flatly for the research register.
| Tool | Engines covered | Prompts at entry | Published entry price |
|---|---|---|---|
| Profound | 9 | 50 | $99/mo · as reported |
| Peec AI | 7 | 50 | ≈$95/mo · as reported |
| Scrunch AI | 8 | 350+ | $300/mo · published |
| AthenaHQ | 11 | ≈3,600 credits | $295/mo · published |
| Evertune | 11 | 100,000 | $800/mo · published |
| Goodie AI | up to 12 | 100 | $399/mo · published |
| Otterly.AI | 4 + paid add-ons | 15 | $29/mo · published |
| PromptRush | 5 | 25 | $19/mo · published |
| Knowatoa | 7 | 30 | $59/mo · published |
| HubSpot AEO | 3 | 25 | $50/mo · published |
Prices are the vendors' published entry tiers as read on their own pricing pages on 2026-09-15; two figures marked “as reported” come from third-party price reporting where the vendor publishes no figure. Self-serve tools in this field cluster between $19 and $800 per month at entry; enterprise suites run higher. Most cover brand monitoring at country-level geography.
Our methodology, for the same register: city-level buyer-question batteries built from measured search terms; ChatGPT, Claude, Gemini (model-only lane), and Perplexity via official APIs; every answer kept verbatim with its sources and date; frequency-based scoring with coverage stated; a gap decomposition against measured site signals; monthly cadence per market. Members receive the full measurement — the battery, the engines, the receipts, the scores, and the trend — at no charge, as part of membership.
Its Own Report, Included With the Free Membership
The AI Answer report is its own report — run separately from the Rank DNA report, on its own allowance, with its own history in your account. Both are included with the same free membership. It runs behind a verified account for the same stated reasons as everything else here: each measurement costs RankAssurance real engine fees to produce, the account prevents abuse of that cost, and your reports are stored to your profile — private, and never re-run just to be re-opened. Every report keeps its own short number for good, so you can find or quote it later without re-running anything — and that same number opens straight into the fixes for that exact measurement, any time, from Your Reports: no searching, no re-running, the workspace waiting where you left it. No card, and nothing to cancel.
Asked Plainly, Answered Plainly
What is AI answer visibility?
Whether — and how prominently — AI assistants name a business when buyers ask them real buying questions. It has four measurable parts: being named at all, the position you are named in, your share of the names an answer market mentions, and whether your own site is cited as a source behind the answers.
Why doesn't ChatGPT recommend my business?
Usually one of a small set of measurable reasons: the assistant has no evidence you exist, its retrieval cannot reach or read your site, there is nothing extractable to quote, your identity is inconsistent across the web, or third-party sources carry nothing about you. Our research piece 'Why Doesn't ChatGPT Recommend My Business?' walks the diagnosis in order.
Can anyone guarantee a business a place in an AI's answer?
No. Answer engines are probabilistic systems that re-decide every answer. What can be done honestly is to measure who is named, how often, and what measurably separates the named from the unnamed. We measure; no one can promise placement, and we never do.
How is AI answer visibility measured?
With a battery of real buyer-intent questions for one market, asked to the answer engines through their official APIs, with every answer kept verbatim. Scores are frequencies across the whole battery — never a single answer presented as a stable rank — and every score carries its coverage: how many questions, how many engines, measured on what date.
Does ranking on page one of Google mean AI assistants will name me?
Not by itself. Classic ranking and answer-engine naming draw on overlapping but different evidence. A site can rank well and still be absent from AI answers — which is exactly why the two are measured separately.
What does the AI Answer report cost a member?
Nothing. It is its own report — run separately from the Rank DNA report, on its own allowance — and both are included with the free membership. Producing it involves real engine fees on our side, which is why it runs on a set cadence per market and behind a verified account.
Related research: Is Your Website Visible to AI? · Why Doesn't ChatGPT Recommend My Business? · What AI Overviews Did to the Click Curve