AI visibility scanner — definition, method, and answer receipt

An AI visibility scanner sends defined prompts to generated-answer systems and records whether the correct business entity is mentioned, recommended, or cited. A scanner result is a dated sample, not a permanent AI rank.

Short answer: A credible scanner preserves one receipt per prompt, engine, and sample: exact question, answer, citations, timestamp, model/mode, entity match, competitors, and provider status. It keeps failures out of the visibility denominator and shows the formula behind every score.

What a scanner measures

Field	Meaning
Mention	The exact target entity appears in the answer
Recommendation	The answer presents the entity as suitable for the stated need
Citation	The answer exposes a source URL
Competitor	Another resolved entity appears in the same answer
Position/prominence	Where and how strongly the answer presents an entity
Availability	The provider returned a valid answer for the planned cell

Keep branded and unbranded prompts separate. “What is [brand]?” measures prompted recognition; “Which [category] provider fits [need]?” measures unbranded discovery.

Minimum answer receipt

{
  "prompt": "Which tools measure AI search visibility?",
  "engine": "ChatGPT",
  "surface": "web search",
  "sample": 1,
  "observedAt": "2026-09-09T11:39:55Z",
  "status": "valid",
  "entityMatch": "reviewed",
  "mentioned": false,
  "recommended": false,
  "competitors": ["example.com"],
  "citedUrls": ["https://example.com/research"],
  "answer": "[complete retained answer]"
}

The VisiScan self-scan CSV publishes 30 real answer receipts. It also exposes a bad OTHER category classification, showing why a scanner must let reviewers inspect prompt fit before trusting the output.

Scanner versus rank tracker

A conventional rank tracker records a page's position for a query in a search-results surface. A visibility scanner records generated prose, entities, recommendations, and citations. Search results and generated answers can coexist and should be measured separately.

Failure and ambiguity rules

• Keep timeouts, blocked responses, and unavailable providers outside the valid visibility denominator.
• Preserve provider coverage as its own metric.
• Resolve similar business names with the canonical domain, location, and context.
• Mark uncertain entity matches for review.
• Deduplicate on prompt-panel version, engine, sample, and run.
• Do not blend live, replayed, and mocked observations without visible provenance.

How VisiScan fits

VisiScan runs buyer questions across available answer engines, extracts mentions, competitors, and citations, and combines those observations with site-readiness checks. Its methodology documents score weighting, confidence, and provenance. The transactional free AI visibility scanner starts a scan; this article explains how to judge the output.

For vendor selection, use how to evaluate an AI visibility scanner. For current vendor descriptions, use best AI visibility tools.

FAQ

Is a visibility score the same as mention rate?

Not necessarily. Mention rate is a numerator divided by valid answers. A score may weight prompt groups or other fields. Require the formula and counts.

How many samples should a scanner run?

Enough to expose the variation relevant to the decision. Two samples are a small baseline; larger samples narrow uncertainty but cost more. Always show the sample count.

Can a scanner explain why an engine omitted a business?

It can identify correlated gaps and cited alternatives. It cannot reveal private model reasoning or prove causation.

Sources

• VisiScan methodology
• VisiScan self-scan data card
• OpenAI crawler documentation
• Google AI feature guidance