AI Visibility Check: What It Measures and Who Audits It

Run a free AI visibility check and the score that comes back looks precise. Then ask the same engine the same buying question ten times — we did, in a dated measurement — and it returns a differently composed provider list on nearly every run. That variance tells you more about AI visibility checks than most tools selling one will.

An AI visibility check measures whether AI assistants mention or cite your brand when they answer the questions your customers ask. A single free check gives you a snapshot. Whether that snapshot means anything depends on two things the score does not show: how often the measurement was repeated, and whether your site is technically able to be cited at all. This guide covers both — and the question of who actually audits this.

Who offers an AI visibility or AI search readiness audit?

The market splits in two: automated checkers from SEO tool vendors — Semrush, Ahrefs and Frase all run free ones — and audit services delivered by agencies, most of them US-based. AI Loopwise covers both ends for the DACH market: a free AI visibility check with results in 2 business days, and a scoped audit covering measurement, crawler access and content substrate.

Nobody in this market, us included, can guarantee AI rankings. What a serious provider can do is measure honestly, name the mechanism behind every finding, and show you the work.

AI visibility check vs. AI visibility audit — what is actually different

A check is a snapshot: a tool probes a handful of AI-shaped queries and reports whether your domain appears. Frase's checker, by its own description, probes three AI-search-style phrases per domain. That is genuinely useful as a first reading, and structurally unable to tell you why you are absent or what to change.

An audit is the layer underneath: who answers your buyers' questions today and from which sources, whether AI crawlers can reach your site at all, whether your content exists in a form an answer engine can lift, and what to build first. A check tells you the temperature. An audit tells you why the room is cold.

What an AI visibility check actually measures

The serious ones measure two different events. A mention is your brand appearing in the answer text. A citation is your page being linked as a source. Engines differ: some answers name brands without linking anything, others cite pages that are never named in the prose. A check that conflates the two overstates whatever it is selling.

The number that gets skipped is run count. AI answers are not deterministic — the same question, asked the same way, returns different providers on different runs. In our own measurement we put each tracked question to ChatGPT and Perplexity five times per engine on every dated run — ten recorded answers per question per measurement day — and log every answer verbatim with every domain it cited. In our 19 and 20 August 2026 runs, ChatGPT assembled a materially different provider list for the audit question above on every single pass — no two of its ten answers matched. Perplexity's cited sources were identical within each day's five runs — between the two days, two source domains swapped out and two in, because it leans on a small set of listicle sources — and those source domains are logged too, because they are where visibility is actually decided. One probe is an anecdote. A run count with logged sources is a measurement.

LLM visibility, GEO audit, AEO — same category, different vocabulary

You will meet this category under several names. LLM visibility is the generic term: are you in the model's answers. GEO (generative engine optimization) is the optimization discipline, so a GEO audit is the same assessment wearing its optimization hat. AEO (answer engine optimization) is the older cousin from the featured-snippet era, now used for AI answers too. There is no meaningful technical difference between an AI visibility audit, a GEO audit and an LLM visibility audit — compare providers on method, not on which acronym they picked.

The readiness layer no checker score shows you

Before any score means anything, the mechanics below decide whether you are even eligible to be cited:

1. Can the crawlers get in? Every answer engine reads the web through named crawlers — GPTBot and OAI-SearchBot (OpenAI), PerplexityBot, ClaudeBot (Anthropic), Google-Extended and CCBot. One inherited robots.txt rule can lock out all of them silently. When we checked 23 large German shops' robots.txt files in August 2026, 19 mentioned no AI crawler at all — most sites have simply never made the decision.
2. Is there a machine-readable layer? Structured data, clean headings, and answer-shaped pages that state the answer before the argument. Engines lift text; text that buries its answer does not get lifted.
3. Skip the llms.txt step you keep reading about. An Ahrefs study of 137,210 domains (May 2026) found 97% of llms.txt files received zero traffic, and no major AI platform commits to reading the file. Google's John Mueller has called it a "temporary crutch", and the one defensible case is a site whose customers use coding agents, which do fetch it. For everyone else it earns nothing measurable — an audit that sells it as a ranking switch is reciting last year's blog posts.
4. Are you in the sources the engines actually use? Answers are assembled from a substrate of pages — directories, comparison posts, documentation. If none of the substrate mentions you, the model has nothing to repeat. This is why visibility work is mostly substrate work, not homepage work.

A zero score with a blocked crawler is a plumbing problem. The same zero with open plumbing is a substrate problem. The fix is different, which is exactly why the score alone is not actionable.

What a zero actually means

If your check comes back at zero, the honest reading is usually undramatic: the engines have not settled on your category's answers yet, or they are assembling answers from sources you are not in. Early categories are exactly where substrate work moves the needle fastest — the position is still open.

We would know: we run our own tracked questions on a fixed schedule, every run in a dated file, and for the audit question at the top of this page our own brand is not in the answers at all yet. We publish that reading rather than hide it, because a measured zero with a dated trail is worth more than a vendor's green badge — it is the baseline every later reading gets judged against.

The part that closes the escape hatch: "not settled yet" does not mean "safe to wait". The substrate sources the engines lean on are being written now, by someone. If you want the honest first reading for your own domain, the check is free and comes back within 2 business days.

Frequently asked questions

Is an AI visibility check free?

The snapshot tier usually is — Semrush, Ahrefs, Frase and we all offer one without payment — ours returns in 2 business days. Paid work starts where the snapshot ends: repeated measurement, source logging, and the audit of why you are absent and what to change first.

What is the difference between AI visibility and traditional SEO?

SEO earns you a ranked position a searcher clicks. AI visibility earns you a place inside a synthesized answer, which may never send a click at all. The overlap is real — crawlability and content quality feed both — but the measurement is different: positions and clicks there, mentions and citations here.

Which AI platforms does an AI visibility check cover?

It varies by tool. The usual set is ChatGPT, Perplexity, Google's AI Overviews and Gemini, and Microsoft Copilot. Ask any provider which engines they probe, how many runs per engine, and whether they log the cited sources — those three answers tell you how seriously to take the score.

How often should I run an AI visibility check?

AI answers move faster than search rankings, so a one-off check ages quickly. We measure our tracked questions on a fixed weekly schedule; for most businesses a reading every 30 days with a consistent question set is enough to see movement without chasing noise.

What is a GEO audit?

The same thing as an AI visibility audit under the generative-engine-optimization label: an assessment of whether AI answer engines can crawl you, whether your content is liftable into answers, and whether you appear in the sources those answers are built from.

Does blocking AI crawlers in robots.txt help or hurt visibility?

For visibility it plainly hurts: a blocked crawler cannot cite you. The real decision is per-bot — you can allow answer-engine crawlers like OAI-SearchBot while blocking training-only crawlers, and that split is exactly the kind of call an audit should make explicit rather than leave to an inherited robots.txt.