AI visibility audit for small businesses
AI visibility can be sampled. It cannot be guaranteed.
This audit records how agreed AI systems respond to a defined set of business questions at a particular time, then checks the public sources and technical conditions behind those responses.
It does not turn changing answers into a permanent rank or pretend that one prompt represents every customer, model, interface, or future response.
- Unit of evidence
- A recorded prompt and response
- Interpretation
- Source-backed and time-bounded
- Excluded
- Fabricated visibility scores and recommendation promises
Define the test before running it
What the baseline records
- The business and question set. Tests begin with confirmed services, locations, audience, and business identity. Brand-name checks and broader category questions are kept separate.
- The systems and conditions. The record names the provider, interface, date, known account or location state, exact prompt, and response. Fresh or signed-out sessions are used where the system permits.
- Mentions, facts, and sources. The audit records whether the business appears, which facts are accurate or wrong, and which sources or links the response exposes.
- Repeatability limits. A small repeat set shows whether an answer is stable within the test window. Variation is reported as variation, not averaged into a false rank.
- Current provider behavior. Each provider's current public documentation and available source behavior are checked at the time of the audit rather than assumed from an earlier test.
The controllable layer
What is checked behind the answers
- Website access and indexing
- Whether important pages are reachable, internally linked, textually clear, and eligible for search indexing.
- First-party business facts
- Whether the website clearly states the real business name, owner, services, locations or service area, contact path, and other facts relevant to the tested questions.
- Public footprint consistency
- Whether the Google Business Profile and other material public sources agree with the website or introduce duplicate, outdated, or contradictory facts.
- Source quality in the observed response
- Whether the answer relies on the business's current sources, an aggregator, an outdated page, no exposed source, or a source that does not support the statement made.
A missing mention does not prove that a system can never find the business. A mention does not prove that the system will repeat it or recommend the business to another person.
Correct the source, not the screenshot
What can be changed honestly
Agreed work can correct crawl barriers, broken links, missing or contradictory business facts, weak internal discovery paths, incomplete profile information, and pages that fail to explain a real service in useful text.
The work does not include invented reviews, fabricated third-party coverage, hidden text, mass-produced location pages, special "AI files" presented as required, or structured data that contradicts the visible page.
Not promised
No AI recommendation, citation, mention, position, permanence, or traffic number. The deliverable is the recorded baseline, the source findings, and verification of agreed source corrections.
A baseline that can be repeated
What the client receives
- The agreed system and prompt matrix with date, conditions, responses, and exposed sources.
- A fact check that separates accurate, missing, unsupported, and contradictory statements.
- A source and technical finding list, with evidence for each material item.
- An agreed correction record and verification that the controllable public source changed.
- A repeat plan that uses the same documented method without claiming the next answer will be identical.