Run a useful AI search visibility audit in one afternoon

Use buyer questions, not a vanity brand check

Asking an assistant whether it knows your company can reveal factual errors, but it says little about how a prospective customer discovers providers. Start with questions customers ask before they know your name: who serves their location, which option fits their situation, what affects cost, and what evidence they should check.

Choose 10 questions from sales calls, support conversations and search data. Include comparison, suitability, local and implementation questions. Preserve the exact wording and market. Keep a separate set of branded accuracy checks rather than mixing them into discovery coverage.

Record what the tool actually showed

Run the same questions across two agreed search-enabled experiences. Record the date, product or model when visible, location, signed-in state, exact prompt, answer, cited URLs, brand mention and any factual error. A mention without a clickable source is different from a citation. A cited competitor is a useful clue, not proof that you need to copy its page.

Repeat the sample under comparable conditions when practical. Answers can vary by wording, retrieval, location and session. Report “mentioned in 4 of 20 sampled answers,” not “20% share of all AI search.” The denominator is your sample, not the entire market. Avoid treating a single answer as a stable rank.

1. Collect 10 commercially relevant questions.
2. Test two agreed experiences and preserve the evidence.
3. Classify mentions, citations, accuracy and competitor examples separately.
4. Inspect the pages cited for those specific questions.
5. Choose one relevant improvement and recheck after it is live.

Fix the information a customer needs

Make the service, audience, geography, scope, pricing factors and next step explicit. Add genuine expertise: a clear process, a sourced explanation, a permissioned case study, a useful comparison or a well-labeled demonstration. Ensure important answers are readable as text and connected through internal links.

Google says ordinary SEO fundamentals remain relevant to its AI features, without special AI markup or extra eligibility requirements. Indexing and citation are not guaranteed. Check crawl access, indexing, canonical URLs and whether structured data matches visible content. Do not sell a special schema package as a shortcut to recommendations.

Connect the audit to a business result

Log the change, its owner and the question it is intended to help answer. Track sampled accuracy and citations alongside observable referrals, qualified enquiries and customer feedback. Visibility may improve before business outcomes, and a lack of tracked referrals does not establish that nobody was influenced.

Authority work must be credible: correct business information, real expert biographies, legitimate reviews and useful participation in relevant communities. Do not manufacture reviews, buy undisclosed endorsements or flood communities with promotional answers. Earn the right to be useful to the audience you want to reach.

Try it with your business

1. Write 10 questions a qualified buyer would ask.
2. Use one row per question and experience, with a citation URL and timestamp.
3. Choose two page improvements and one technical check.
4. Schedule a monthly repeat and retain the original question set.

Use the planning worksheet and save it as a PDF →

Use AI to help organize the work

Provide only material you are allowed to use. This prompt helps structure your evidence; it does not give the tool access to your analytics or verify facts on its own.

Turn these actual customer questions into a 10-question AI discovery audit for [service/market]. Keep branded accuracy checks separate. Create a recording table for date, experience, context, exact question, mention, citation URL, accuracy and uncertainty. Do not claim to have run the audit or inspected answers.

Measure: Mentions, citations and accuracy within a defined sample, plus qualified enquiries where observed.

Sources & further reading

Official references checked October 26, 2026. Product reporting and requirements can change.

• Google Search Central: AI features and your website