EST. MMXXVI · THE INSTITUTION OF RECORD FOR AI CITATION · OPEN METHODOLOGY

The AI Citation Institute sealTHE AI CITATION INSTITUTEEx corpore, auctoritas.

The AI Citation Institute · The Answers

What does an AI visibility audit include?

An AI visibility audit measures whether AI answer engines name your business when real buyers ask, and it covers four things. First is a baseline probe panel: the actual questions your buyers type, run across a frozen five-engine panel (ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews), each query sampled k times and scored mention, recommendation, or absent — which produces a gap map of exactly which queries name a competitor instead of you. Second is a retrieval and rendering check: whether your pages are server-rendered and retrievable at all, whether schema is present, missing, or set to the wrong vertical, and your robots, llms.txt, and canonical hygiene. Third is internal-link topology, because inbound internal links are the strongest on-page lever we measure — in a 91-post audit, pages with zero inbound links were self-cited about 4.5% of the time, and pages with ten or more hit roughly 44%. Fourth is competitor context: how rivals score on those same queries, so a gap reads as a specific business someone recommends over you rather than a vague blank. The output is a scored panel, a gap map, and a prioritized fix list.

The four components

A real audit mirrors the free-scan deliverable. Each component answers a different question, and together they explain not just whether AI recommends you but why.

What each audit component covers and what it reveals
ComponentWhat it reveals
Baseline probe panelYour real buyer questions run across the frozen five-engine panel, each sampled k times and scored mention / recommendation / absent — a gap map of which queries name competitors instead of you.
Retrieval & renderingWhether pages are server-rendered and retrievable, whether schema is present, missing, or wrong-vertical, plus robots, llms.txt, and canonical hygiene. A page an engine cannot read cannot be cited.
Internal-link topologyHow your pages link to each other — the strongest on-page lever we measure. Orphaned pages rarely get cited; well-linked pages do.
Competitor contextHow rivals score on the same queries, so each gap reads as a specific business someone recommends over you, not a blank.

Why the probe panel is the core

The panel is the instrument. It is frozen so results are comparable run to run: the same five engines (ChatGPT, Claude, Gemini, Perplexity, Google AI Overviews), the same buyer questions, each sampled k times because a single answer is noisy. Sampling several times turns a coin-flip into a rate.

Every query gets scored mention, recommendation, or absent. Roll those up and you get a per-engine visibility rate plus a gap map: the specific questions where an engine names a competitor and never you. That map is what the rest of the audit exists to explain and the fix list exists to close.

k = the number of times each query is sampled per engine. Rates, not single answers.

Why retrieval and links decide the outcome

An engine can only cite a page it can retrieve. The rendering check catches the silent failures: content that only exists after client-side JavaScript, missing or wrong-vertical schema, a robots or canonical setup that hides the page. These are cheap to fix and gate everything else.

Internal links are the strongest on-page lever we have measured. In a 91-post audit, pages with zero inbound internal links were self-cited about 4.5% of the time; pages with ten or more inbound links hit roughly 44%. Topology is why two pages with similar content get cited at wildly different rates.

Common questions

How is an audit different from the free scan?+

The free scan runs a slice of the probe panel on your domain so you can see the gap map on a live engine. A full audit adds the retrieval and rendering check, internal-link topology, competitor context on the same queries, and a prioritized fix list.

Which engines does the panel cover?+

A frozen five-engine panel: ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. It is frozen so scores are comparable from one run to the next.

Why sample each query more than once?+

A single AI answer is noisy — the same question can name a competitor one time and you the next. Sampling each query k times turns that into a stable rate you can track and improve.

keep reading

See what AI says about your brand right now — free, on a live engine.

Scan your domain