An AI visibility audit is a structured observation of how selected AI discovery systems respond to commercially relevant questions. It is useful when it records evidence and limitations. It becomes misleading when it turns a small sample into a permanent “rank.”
What the audit should measure
| Field | Record | Why it matters |
|---|---|---|
| Question | Exact wording and buyer stage | Keeps future runs comparable |
| Conditions | Engine, date, market, language and account state | Explains why answers may differ |
| Brand outcome | Mentioned, recommended, absent or unclear | Separates presence from endorsement |
| Position | Order and surrounding alternatives | Shows competitive context without claiming a universal rank |
| Sources | URLs, domains and source type | Connects the answer to public evidence |
| Confidence note | Ambiguity, instability and verification gaps | Prevents overconfident conclusions |
The seven-step method
Define the decision before opening an AI tool.
Choose one business, market and decision. Examples include understanding which competitors appear for evaluation questions or finding public evidence gaps before a product launch.
Build a balanced buyer-question set.
Cover awareness, consideration, evaluation, trust and purchase. Avoid fifty near-duplicate prompts designed to manufacture a favorable percentage. A small fixed set is more useful than a large moving target. Use the free prompt generator to draft twelve questions, then approve and freeze the final wording.
Freeze the test conditions.
Record engine, date, language, market and relevant account or personalization state. Use a fresh conversation where appropriate and preserve the exact prompt wording.
Capture the complete answer and its sources.
Record the answer, named brands, recommendation order, caveats and cited URLs. A screenshot can support the record but should not replace searchable text and source links.
Separate observation from interpretation.
“The brand was absent in this run” is an observation. “The brand is invisible everywhere” is an unsupported generalization. Keep those statements in different fields.
Audit the public evidence behind the answer.
Check whether important pages are crawlable, internally linked, factually consistent and supported by primary or credible third-party evidence. Google says normal SEO fundamentals remain relevant to its AI features, including crawl access, internal links, textual content and matching structured data. OpenAI says sites that want discovery in ChatGPT search should not block OAI-SearchBot.
Prioritize actions and schedule a comparable rerun.
Convert evidence gaps into owned actions with dates. Repeat the same baseline only after enough time or meaningful changes have occurred. Report direction and evidence, not guaranteed causation.
Four mistakes to avoid
- Combining different engines and dates into one unexplained score.
- Counting a brand mention as a recommendation or a citation as proof of quality.
- Changing prompts between runs and calling the difference an improvement.
- Publishing automated conclusions without checking the underlying answer and source.
Official guidance used
Google Search Central: AI features and your website — crawlability, index eligibility, internal links, textual content, structured-data consistency and measurement.
Google Search Central: optimizing for generative AI features — people-first, non-commodity content and foundational SEO rather than special “GEO hacks.”
OpenAI: Publishers and Developers FAQ — OAI-SearchBot access, referral tracking and the distinction between search discovery and GPTBot controls.
Can your current process produce reviewable evidence?
Score source evidence, repeatable measurement and commercial delivery in five minutes. Answers stay in your browser and no signup is required. The site records only an anonymous CTA click and completion count.
Start the free assessmentFrequently asked questions
What does an AI visibility audit measure?
It records how selected systems answer a fixed set of buyer questions, including mentions, recommendation position, cited sources and competitors under documented conditions.
Is AI visibility a permanent ranking?
No. Answers can vary by system, model, account, location, language and time. The audit is a dated evidence baseline.
How often should the audit be repeated?
Repeat only when the question set and conditions are stable enough to compare. Monthly or quarterly can be useful depending on the business and decision cycle.