Diagnosis

AI visibility checker: what a check actually is, and how not to fool yourself.

A checker asks AI assistants questions about your business and reports who they name. That is the whole mechanism. The differences are which questions, which engines, whether you see the actual answers — and whether one snapshot gets sold to you as a trend. What to expect from a check, how to read one honestly, and where to run a free one.

Updated 2026-10-02 · Prices and data verified 2026-08-11

Every AI visibility checker does the same thing: it asks assistants a set of questions and records who they name. Everything that differs between tools — and the differences are real — sits on top of that one mechanism. So the useful question is not “which checker” but “what does a check consist of, what can a free one tell me, and how do I read the result without fooling myself.” This page answers those, in that order, and then points at ours.

What a check consists of

Four observations, per question, per engine: mentioned (was your business named at all), linked (was your site cited), where (how early in the answer you appeared), and who else (which competitors were named). A checker runs those across a question set and an engine set, and then either shows you the answers or compresses them into a score. Our metric page prints the formula we use for the compression; this page is about the observations, because they are the thing.

What the free check here reports

Your three numbers. How often the engines mentioned you, linked to your site, and named you first, measured across 10 buyer questions asked live on ChatGPT, Gemini, and Perplexity (30 queries in all).
The answers themselves, word for word. For the three questions that expose the most, exactly what each engine wrote, with your name and every competitor's highlighted in the text.
Where you stand against the names it picked instead. A plain-English read of which competitors the engines named, and how big the gap is.
Up to 5 worked examples. Real questions from your own scan: what the AI did with each one, why it happened, the specific fix, and roughly what it takes.
Three moves, in order. The first work order, the one for this week, and how the work is tracked after that.
A PDF that's yours for good. Download it and it's permanent, and the link to it opens without an account. The audit asks for your email so a link to the report reaches your inbox too; there is nothing else to fill in and no card at any point.

Seven ways a check fools you — our list

One question. Engines compose fresh answers each time; a single question can name you or not by luck. Read across several.
Your brand name as the question. Of course the engine knows your name when asked for it. Buyers ask about their problem, not your brand. Check the problem questions.
One engine. A business can own the ChatGPT answer and be absent from Perplexity on the same question. A check on one surface is a check on one surface.
A score with no answers behind it. A number you cannot trace to a specific answer is a number you are taking on faith. Insist on seeing the verbatim text.
A model asked what it believes. Asking a model what it thinks of a brand is not a live answer a buyer would see. A check that blends the two without saying so is measuring two things and reporting one.
A snapshot sold as a trend. 40–60% of the sources AI cites change every month (eMarketer, 2026). One check is where you stood on a day.
Comparing scores across tools. Different questions, engines and weights make different measurements. The only comparison that holds is your own, over time, on the same setup.

Other checkers, as our verified dataset describes them

Two monitoring products from established SEO suites are documented in the dataset behind this site, and stated here exactly as it states them (prices checked August 2026):

Semrush AI Visibility Toolkit. Monitors only: $99/mo per domain for 25 prompts across 3 engines; reviewers note the 'Growth Actions' execution feature was still listed as coming soon. No content generation or CMS publishing in the toolkit.
Ahrefs Brand Radar. Monitors only: add-on tiers $50–699/mo on top of an Ahrefs subscription (independent analyses put realistic full coverage at $828–1,148/mo). No Claude coverage, no content production layer, no GA4 attribution.

Both are legitimate ways to get the number. Neither shows the fix or performs it, which is a description of scope rather than a criticism, and the wider category — priced on this page — runs $29–495/month.

Reading the result

Absent from the answers.The engine cannot place you. This is an entity problem before it is a content problem: the same name, description and category everywhere, and a few pages that answer a buyer's question in plain text.

It depends

Present but outnumbered.The engine lists a crowd on the questions that matter. The work is contesting the winnable ones — the specific, local, long-tail questions where a clear page beats a national brand's generic one.

Yes

Present but late.Someone else's page is the source and yours is an also-mentioned. The work is becoming the page the answer is built from: the cost drivers, the specifics, the quotable sentence. This is the best position to start from.

08 · FAQ

The questions people ask about checkers.

What is an AI visibility checker?

A tool that asks AI assistants a set of questions about your business or category and reports what came back: whether you were mentioned, whether your site was linked, where you appeared in the answer, and who was named instead. That is the whole mechanism. The differences between checkers are in which questions they ask, which engines they ask, how they read each engine, how often they re-ask, and whether they show you the actual answers or only a number derived from them. The last difference is the one to care about most.

Is a free check enough?

For the first question — does the engine know you exist, and who does it name instead — yes, usually. Our free audit runs 10 buyer questions across ChatGPT, Gemini and Perplexity, thirty live answers, three of them shown word for word with every name highlighted, plus a standings read, worked examples and three moves to make, in order, with no account and no card, just an email for the report. What a free check is not is a trend: it is one snapshot. Whether anything you changed moved the answer needs the same questions asked again later, which is what a paid scan on a schedule is for.

What other checkers exist?

Two documented in our verified dataset (prices checked August 2026), stated as the dataset states them. Semrush AI Visibility Toolkit: Monitors only: $99/mo per domain for 25 prompts across 3 engines; reviewers note the 'Growth Actions' execution feature was still listed as coming soon. No content generation or CMS publishing in the toolkit. Ahrefs Brand Radar: Monitors only: add-on tiers $50–699/mo on top of an Ahrefs subscription (independent analyses put realistic full coverage at $828–1,148/mo). No Claude coverage, no content production layer, no GA4 attribution. Both are monitoring products from established SEO suites, and both are legitimate ways to get the number; neither shows the fix or does it. The wider category runs $29–495/month and is priced on the pricing page linked below.

Why do I get a different result each time I check?

Because the engines do. Assistants compose a fresh answer on every ask, retrieval changes as the web changes, and the same question phrased two ways can name two different shortlists. A single check can therefore under- or over-state you by luck. The honest way to read any checker is across several questions rather than one, and across time rather than once. A tool that reports a single-question, single-run number as your 'score' is reporting noise with confidence.

What should the result make me do?

Read the verbatim answers before the number. If you are absent, the problem is usually entity — the engine cannot place you — and the fix is consistency and a few question-shaped pages. If you are present but outnumbered, the problem is the questions where the engine lists a crowd, and the fix is contesting the winnable ones. If you are present but late in the answer, the problem is that someone else's page is the source and yours is an also-mentioned. The number cannot tell you which of those you are; three answers, word for word, can.

How do I check without a tool at all?

Open ChatGPT, Gemini and Perplexity and ask each one five questions a real buyer would ask — not your brand name, the problem your buyer has — and write down who gets named. It takes twenty minutes and it is what every checker automates. The tool earns its place when you want the same questions re-asked on a schedule, the answers stored, and the difference shown. Until then, twenty minutes and a notebook is a perfectly honest AI visibility check.

Keep reading

AEO fundamentals

AI visibility

The metric defined, and the thing vendor pages avoid: the formula, printed from the code that runs it.

Tool comparisons

Best AI visibility tools

The stage-matrix listicle: who monitors, who diagnoses, who executes, who proves — honestly placed.

Diagnosis

Not showing up in ChatGPT

The six real reasons — and a ten-minute self-diagnostic to find out which one is yours.

Diagnosis

Does ChatGPT know you

The five-prompt manual check, the three outcomes, and why a single check can mislead you.

Run the free check: 30 live answers, three shown word for word.

Ten buyer questions across ChatGPT, Gemini, and Perplexity, about two minutes, every name highlighted. A permanent PDF, just an email, no account, no card.

Run the free audit Start the 7-day trial

Free audit first: see exactly where you stand before paying anything. 7-day trial, no card.