AI Visibility Audit: How to Check What AI Says About Your Business

An AI visibility audit checks how selected AI systems describe your business, which sources they cite, what facts are missing or wrong and whether the answers change over time. Start with a small fixed set of customer questions, run them under recorded conditions, capture the exact answers and sources, then separate factual errors from ordinary differences in wording. The goal is not to force one answer. It is to create a repeatable record that tells you what needs attention.

Define what you are trying to learn

AI visibility is broader than whether your business name appears once. A useful audit identifies the decision behind each question. A customer might be checking whether you serve a location, whether you offer a particular service, who leads the company or which providers seem relevant to a problem. Each question should connect to a real fact or choice that matters to the business.

Choose one business, one audience and one review period. Do not mix the founder, the company and several services into a single score. That makes it hard to see which source or identity needs correction. Reputation Repair’s Google and AI reputation service uses separate observations for search results and AI answers because they measure different things.

Build a fixed prompt set

Use a small group of prompts that you can repeat. These are test questions, not proof of what every customer asks. Keep the wording stable for comparisons, then add a few natural variations to see whether the description changes when the question changes.

• What does this business do?
• Who does this business serve?
• Where is this business located and where does it operate?
• Who leads the company?
• What services does the company offer?
• What sources explain the company’s experience?
• Which companies help with the problem the business is trying to solve?
• What should a buyer verify before choosing a provider in this category?

Add service and location questions only when they reflect the real market. A Gilbert business serving clients nationwide might test one local question and one national question. Do not add cities merely to create keyword variations. The prompt should represent a customer decision the company can actually support.

Record the conditions for every answer

AI answers can vary between runs. Record the system, date, model when shown, whether web search was used, account or personalization state when relevant, exact prompt, exact answer and visible sources. OpenAI explains that ChatGPT responses using web search may include citations and a Sources view. Google advises users to check important information in more than one place because AI Overviews can make mistakes.

Run the baseline in a clean and consistent way. If you later change the prompt, system or settings, label it as a new test instead of treating it as a direct comparison. A screenshot is useful, but the written record should also preserve the answer text and source links so another reviewer can verify what was observed.

Score five different signals separately

A single visibility score can hide the most important finding. Review each answer across five distinct signals.

1. Mention: The business or person appears in the answer.
2. Citation: A page from the business or another relevant source is linked.
3. Description: The answer states accurate services, roles, locations and context.
4. Recommendation: The business is presented as an option for the customer’s question.
5. Action: A measurable visit, inquiry or conversion occurs after the exposure.

These signals are not interchangeable. A citation is not automatically a recommendation. A mention is not a website visit. A visit is not a qualified inquiry. A changed answer does not prove that a campaign caused the change. Keeping these stages separate prevents an encouraging observation from becoming an unsupported performance claim.

Classify every issue before acting

Not every difference requires a correction. Use four categories to decide what happens next.

• Accurate and useful: Keep the evidence and repeat the observation later.
• Accurate but incomplete: Identify the missing fact and the public source that should explain it.
• Incorrect with a traceable source: Correct the owned page or pursue an appropriate update from the publisher.
• Incorrect without a clear source: Preserve the answer, test related prompts and look for conflicting public information before choosing an action.

Prioritize errors that can affect a buyer’s decision, such as the wrong business category, former leadership listed as current, an outdated address or services the company does not provide. Minor wording differences usually matter less than a false fact or a confusing identity match.

Strengthen the public sources behind the answer

Start with sources the business controls. Confirm that the website clearly states the company name, services, leadership, location, contact path and important qualifications. Align major public profiles with those facts. If structured data is used, it must match the visible page. Google’s structured data guidelines require markup to be accurate, current and representative of the page rather than hidden or misleading.

Then review outside sources that the selected system cites. A correction request may be appropriate when a publisher page contains an objective error. New original content may help when the public record lacks useful context. Neither action guarantees an AI citation or a particular answer. The practical aim is accurate, accessible source material that can also help human readers.

Use a repeatable comparison schedule

Repeat the same core prompts on an agreed cadence. Compare answer descriptions, cited sources, material errors and whether the business appears when the question asks for providers. Keep newly published work separate from confirmed changes in AI answers. A page can be published without being cited, and an answer can change without demonstrating why.

A simple report should show the baseline, current observation, exact source, completed work, unresolved concern and next owner. Reputation Repair’s monitoring service records agreed searches and selected AI answers as samples rather than a complete picture of everything every user sees. The reporting example also separates completed work from observed change and the next decision.

A practical thirty minute audit

1. Choose three customer questions and two factual questions.
2. Run each prompt in the selected system and record the conditions.
3. Save the answer and every visible citation.
4. Mark mentions, citations, descriptions and recommendations separately.
5. Verify material facts against current public sources.
6. Assign each issue to accurate, incomplete, traceable error or unclear source.
7. Choose one correction or content action with a named owner.
8. Set the next comparison date without promising a result.

This short audit will not measure the entire market, but it can reveal a wrong address, outdated role, unclear service description or missing source that deserves action. It also gives the business a baseline it can compare honestly.

When outside help makes sense

A business can run a limited audit internally when the questions, sources and owners are straightforward. Outside support may help when several identities overlap, damaging information is involved, sources conflict or the team needs ongoing monitoring and coordinated correction work. Reputation Repair scopes Google and AI monitoring according to the names, queries, prompts and review frequency involved. Broader Google and AI reputation work is provided through a custom proposal, with shared research and content listed once rather than charged twice.

The best AI visibility program begins with disciplined observation. Record what the system said, verify the sources, correct important facts and repeat the same test later. That creates useful evidence without pretending that one answer represents every platform, every user or every future result. Review the current services and pricing if you need help defining the systems, questions and reporting scope.

Common questions

What is an AI visibility audit?

It is a documented review of how selected AI systems describe a person or business, including mentions, cited sources, material facts, recommendations and changes across repeat tests.

Can a business guarantee that an AI system will cite its website?

No. A business can improve accurate source material and monitor observations, but it cannot guarantee a citation, recommendation or specific generated answer.

How often should AI visibility be checked?

Choose a cadence that matches the business risk and how often relevant sources change. Use the same core prompts and recorded conditions so later observations are comparable.