What should an AI visibility audit include?

An AI visibility audit should define the questions, providers, market, dates, and competitors; preserve the answers; separate mentions from recommendations and citations; check accuracy and sources; and turn the clearest gaps into prioritized client actions. The report should also explain its limits.

1. Write the scope before running prompts

Agree on the business units, products, locations, languages, competitors, and buying stages included. List the AI services that will be checked and the collection dates.

State what is not included. A small baseline may cover thirty prompts, three providers, one country, and five competitors. A larger international program requires different pricing and review time.

This prevents the audit from turning into unlimited research.

2. Build the question set

Use customer language from sales, support, search data, reviews, and interviews. Group questions into discovery, suitability, comparisons, cost, location, trust, risk, and post-purchase concerns.

Label branded and unbranded prompts. Add the audience, buying stage, and reason for each question. Keep a stable core set so a future audit can make a fair comparison.

PROMPT WORKSHEET
  1. Exact customer question
  2. Branded or unbranded
  3. Audience and buying stage
  4. Market or location
  5. Why the question matters

3. Capture the evidence

For every check, save the provider, available model information, date, question, full answer where permitted, brands, recommendation language, visible citations, and completion status.

Keep failed or unavailable checks separate from valid answers without a mention. Preserve screenshots or exports only when the provider's terms and the client's data rules allow it.

4. Classify each observation

  • Recommended: presented as a suitable choice.
  • Mentioned: named but not clearly recommended.
  • Cited: a page or source was shown.
  • Accurate: important business facts were correct.
  • Ambiguous: the answer may refer to another entity.
  • Missing: absent from a completed answer where relevant competitors appeared.

Do not combine these labels into a score until the client can see how the score is calculated.

5. Review the sources behind the answers

Group visible citations by source type: company website, competitor website, directory, review site, publication, social profile, public document, or other source.

Check whether the client's pages appear, whether important outside sources contain old information, and which domains repeatedly support competitor recommendations.

A source gap does not automatically mean the client needs backlinks. The right action may be correcting a profile, publishing a missing explanation, contributing useful research, or earning relevant coverage.

6. Check description accuracy

Review names, products, services, locations, audience, price language, integrations, ownership, and plan availability. Mark each issue by customer impact and identify the likely source.

Correct what the client controls first. For outside sources, decide whether the information can be updated through a profile, publisher request, or new authoritative material.

7. Compare competitors fairly

Show which competitors appeared for which question groups and the reason stated in the answer. Avoid claiming access to a provider's private ranking formula.

Separate a competitor that appears once from one that appears across several valuable prompts. Look for observable differences in category clarity, useful pages, public reviews, third-party coverage, and consistency.

8. Turn findings into client actions

Group work into technical access, entity and business facts, content gaps, local profiles, reputation, digital PR, and measurement.

For each recommended action, include the original evidence, expected customer benefit, effort, owner, due date, and review method. Use plain language: “Update the pricing page so agencies can see which plan includes client reports” is more useful than “improve entity context.”

ACTION RECORD
  1. Observed question and answer
  2. Customer or business impact
  3. Recommended change
  4. Owner and due date
  5. Evidence to review later

9. Use transparent scoring

If the agency uses a score, publish the formula. Show the number of valid checks, providers, prompts, markets, and dates beside it. Avoid comparing clients whose audits use different samples.

A simple score can help executives scan the report, but it must never hide the original answers.

10. Present the report in decision order

  1. Executive summary in everyday words.
  2. Audit scope and limitations.
  3. Important brand and competitor observations.
  4. Description accuracy issues.
  5. Citation and source patterns.
  6. Work completed since the last review.
  7. Prioritized actions and owners.
  8. Next review date.

Use the broader AI search visibility reporting guide when building the recurring client report.

Make the audit repeatable

Save the prompt set and collection method. Record changes to providers, markets, or scoring. Repeat the baseline after meaningful work or on an agreed schedule.

VisibleGen connects AI visibility evidence, competitor observations, source opportunities, tasks, owners, and client reporting. The platform can organize the work; the agency remains responsible for interpretation and truthful client communication.