An AI visibility audit answers four questions in order: can AI systems reach your site, can they understand it, do other sources corroborate you, and do the answers actually name you. The bottom line: run it in five layers, access, content, entity, corroboration, and outcomes, score each with evidence you can show, and finish with a prioritized fix list and a baseline you can repeat.

The five layers

LayerWhat you checkHow
1. AccessSearch crawlers are allowed and pages are fetchable and indexedrobots.txt for OAI-SearchBot, Claude-SearchBot, and PerplexityBot, CDN and firewall rules, Bing and Google indexing
2. ContentContent is in the raw HTML, specific, structured, and currentcurl the page, count raw words, check headings, answer first structure, dates
3. EntityOne consistent business identityName, address, phone, and description across site, profiles, and directories, plus Organization or LocalBusiness markup
4. CorroborationOther sources name and describe youReviews, directories, press, community mentions, and which sources the answers cite
5. OutcomesAI answers name you20 to 50 customer questions, five runs each, on ChatGPT, Gemini, and Perplexity

Steps

  1. Run the free scan on your homepage and one service page. It checks crawler access for 16 AI bots, llms.txt, schema, server rendered content, meta tags, content density, headings, and answer ready content.
  2. Read robots.txt. Confirm the search and user fetch bots are not blocked. Blocking training bots is a separate decision.
  3. Check raw HTML. Fetch key pages with curl. Under 30 words in the raw HTML is an empty shell, and under 450 is thin by our scan's thresholds.
  4. Verify the site in Bing Webmaster Tools and Search Console. Check indexing and open the AI reports.
  5. Audit entity consistency. List every profile and compare name, address, phone, and description.
  6. Audit corroboration. Count reviews by platform, list directories, and note press and forum mentions.
  7. Run the outcome test. 20 to 50 questions, five runs each, on three platforms, and record shares for you and competitors.
  8. Write the findings with evidence. Each finding needs a screenshot, a log line, or a count.
  9. Prioritize. Access problems first, then content gaps for missed questions, then corroboration.

What the audit should not claim

What a deliverable looks like

SectionContents
BaselineShare of runs naming you and each competitor, per platform and question
BlockersTechnical issues with evidence and the fix for each
GapsQuestions where you are absent, and what the named businesses have that you lack
SourcesDomains cited across answers, ranked, with your presence on each
PlanFixes ordered by effort and effect, with a 90 day timeline

One published guide puts one time AI visibility audits between $2,500 and $10,000, which reflects how much evidence gathering the outcome test alone requires.

Questions people ask

What is an AI visibility audit?

A structured review of whether AI systems can reach and understand your site, whether other sources corroborate you, and whether AI answers actually name you, ending in a prioritized fix list and a baseline.

What should an AI visibility audit include?

Five layers: crawler access and indexing, content, entity consistency, corroboration on outside sources, and an outcome test of customer questions run many times on several platforms.

How long does an audit take?

A basic technical review takes hours. The outcome test of 20 to 50 questions with five runs each on three platforms takes several more hours to run and record.

Does a good audit score mean AI will recommend me?

No. Readiness scores measure inputs. Only repeated runs of customer questions show whether AI names you.

Sources