← Back to the blog

PIP BLOG

What should an AI visibility audit include?

By Pip · Published October 20, 2026

A useful AI visibility audit shows which customer questions were tested, whether your business appeared, what the answers said and which sources were linked. It should also check the information behind those answers and explain what to improve first. A score without its method and evidence is not enough.

Yes, someone can check whether AI tools recommend your business. The important question is what they test and what you receive afterward. A screenshot of one answer is a snapshot, not a complete assessment.

Start with a clearly defined scope

Agree on the business location or service area, services, languages and AI tools being checked. A report about one city should not imply it covers your entire market. A ChatGPT check should not be presented as a test of every AI platform.

Ask whether the work includes recommendations only or also implementation. Checking a profile and correcting it are different tasks.

1. Questions that reflect real customer needs

The audit should list the exact prompts. Include discovery questions that do not name your business and separate questions that check what AI says about your brand.

These are illustrative prompts, not verified popular searches. Choose questions from your actual services and customer conversations.

2. Evidence you can inspect

For each check, request the date, tool, exact prompt, location context, answer and available source links. Record whether web search was used. Keep observations separate from explanations: “your business was not mentioned” is observable; “you were excluded because of your reviews” needs evidence.

Report fieldWhy it matters
Prompt and contextShows what was actually tested.
Date and toolMakes later comparisons more useful.
Mention and descriptionSeparates being named from being described accurately.
Cited sourcesLets you inspect the linked information.
Next actionConnects a finding to work you can do.

3. A review of your public business information

Look for checks of your service pages, contact details, opening hours, location and relevant profiles. The report should identify specific URLs and mismatches, rather than simply saying “improve your online presence.”

A technical review can also check whether important pages are publicly accessible and whether crawler settings need attention. OpenAI describes its search crawler separately from its training crawler in its crawler documentation.

4. A prioritized action plan

Each recommendation should say what to change, where to change it and who will do the work. Ask the provider to distinguish confirmed errors from opportunities they suggest testing.

Illustrative finding

Your website lists Saturday appointments, but one business profile says you are closed. Confirm the real schedule, correct the inaccurate source and record the change.

This example is not a finding from a client audit.

5. A repeatable way to check progress

Keep a consistent group of questions for future checks. If a report includes a visibility percentage, ask for the calculation and number of tests. For example, appearing in 3 of 10 tested answers is a result for that sample, not a claim that you reach 30% of customers.

Track website visits and successful inquiries separately. More mentions do not automatically mean more bookings. Answers can vary, so a report should state its limitations and avoid guaranteed placement.

What to ask before you agree

An initial public-information check does not require you to share your email password. Broader analytics or implementation work may need separately agreed access.

Keep reading