What an AI visibility audit covers
A useful audit answers four questions, in this order:
- Can AI crawlers read your site? If robots.txt blocks them or your text only appears after JavaScript runs, nothing else matters much.
- Do AI answers mention you? Measured as a rate across many buyer-style questions and repeated runs, not a single answer.
- Who gets recommended instead? The competitors named most often are the ones the models associate with your category.
- Which pages do the answers cite? Listicles, Reddit threads, review sites and directories. This is the part you can act on.
Step 1: technical readiness
Check robots.txt for the AI crawlers: GPTBot (OpenAI training), OAI-SearchBot (ChatGPT search), ChatGPT-User (pages a user asks ChatGPT to open), Google-Extended (Gemini), PerplexityBot and ClaudeBot. Blocking a training crawler is a legitimate choice; blocking a search crawler such as OAI-SearchBot usually means you can’t be cited in those search answers.
Then view your page source (not the rendered page). If your product description isn’t in the HTML, crawlers that don’t run JavaScript see an empty shell. Add a clear title, a meta description that says what you do and for whom, Organization or SoftwareApplication JSON-LD, a sitemap.xml and, optionally, an llms.txt summary.
Step 2: write buyer questions
Use the words buyers use, not your tagline. Mix “best X for Y”, “alternatives to [category leader]”, “X vs Y”, “how do I…” and budget questions. Fifteen questions across 3–5 category keywords is enough to see a pattern; the full report uses 15.
Step 3: ask more than once, on more than one engine
The same question can get a different answer a minute later. Asking each question 2 times on OpenAI and Gemini turns noisy single answers into a mention rate. Note that API answers differ from the consumer apps, so read rates as a direction, not an exact share.
Step 4: list the sources and act on them
Group every cited URL by type and work from the most cited down:
- “Best X” articles and listicles: email the author or editor with a two-line pitch and a reason your product fits the list.
- Reddit and forums: answer the actual question, disclose that you work on the product, and follow the community rules.
- Review sites and directories: submit or claim your listing, complete the category and feature fields, and ask real customers for reviews.
- Competitor pages: note which of their pages get cited (comparisons, pricing, docs) and cover the same questions honestly on your site.
Step 5: re-check
Outreach and listings take weeks to be crawled and picked up. Re-run the audit after about a month and compare mention rates and sources. The full report includes one re-check after 30 days.
Free vs full audit
The free AI visibility checker gives you the readiness checks and a 5-question preview on one engine. The full AI visibility audit report runs 15 questions on both engines, 2 times each, for a one-time $9. For a single score instead of a to-do list, see AI visibility score; for the generative engine optimization angle, see the GEO audit guide.
AI visibility audit FAQ
›How long does an AI visibility audit take?
The automated part takes a minute or two for the free preview and a few minutes for the full report. Working through the source list (contacting authors, claiming listings, answering threads) is the slow part and is spread over weeks.
›How often should I run an AI visibility audit?
Monthly is a sensible rhythm. AI answers and the pages they cite change, and outreach takes time to show up. The full report includes one re-check after 30 days for that reason.
›Can an audit guarantee that AI will recommend me?
No. Nobody controls what a model says. An audit shows where you stand on a sample of questions and which pages are worth working on; the outcome still depends on the models and on those pages.
›What's the difference between an AI visibility audit and an SEO audit?
An SEO audit looks at how search engines rank your pages. An AI visibility audit looks at whether AI assistants mention you in their answers and which third-party pages they rely on. The crawler and HTML checks overlap; the rest is different.