AI Strategy
How to Run an AI Visibility Audit on Your Business (Free, 30 Minutes)
Updated Muhammad Taha7 min read
An AI visibility audit is a simple test of what AI assistants and AI search say when your customers ask about your category. You write 15 to 25 realistic questions, ask each of them in ChatGPT, Gemini, Perplexity and Google, and record whether your business is named, how it is described, which competitors appear and which sources are cited. It takes about 30 minutes for a first pass and needs no paid tools.
Key takeaways
- You are measuring three things: are you mentioned, is it accurate, and who is mentioned instead.
- Use the same questions every month so you can compare. Change the wording and the results change.
- Answers vary between runs, accounts and locations. Run each important question twice and note patterns, not single results.
- The cited sources tell you where to focus: they are the pages the system trusts for your topic.
- Save the full text of answers. They are your evidence if something is wrong.
Step 1: Write your question list
Think like a customer, not like a marketer. Cover these five types, with three to five questions each.
| Type | Example for an accounting firm in Karachi |
|---|---|
| Category discovery | "Best accounting firms in Karachi for small businesses" |
| Problem or need | "How do I register a small company and handle tax in Pakistan?" |
| Price and process | "How much does outsourced bookkeeping cost in Pakistan?" |
| Comparison | "Outsourced accountant vs in-house accountant for a 10-person company" |
| Brand check | "Is [your firm] reliable?" and "Who runs [your firm]?" |
Step 2: Set up a clean test
- Use a private window or a logged-out session where possible, so personal history has less influence.
- Note the date, the tool and, for tools with modes, whether web search was on.
- If your customers are in a specific city or country, set your location or add it to the question. Local results differ.
- Ask each question in at least ChatGPT, Gemini, Perplexity and Google. Add Claude or Copilot if your customers use them.
Step 3: Record what you see
Copy this into a spreadsheet, one row per question per tool:
Date | Tool | Question | Mentioned? (Y/N) | Position (1st, 2nd, ... or none)
Description of us (copy the text) | Accurate? (Y / partly / N)
Competitors named | Sources cited (URLs) | Notes
Keep the full answer text in a second column or tab. When you re-run the audit, you will want to compare wording as well as presence.
Step 4: Read the results
Look for patterns across the whole sheet, not individual wins.
- Never mentioned: either the systems do not know you, or your content does not answer these questions. Check whether competitors who are named have clear pages for the same questions.
- Mentioned but wrong: old address, retired service, wrong founder, wrong city. Find the source it used and fix that page, then your own site.
- Mentioned for the wrong thing: you are seen as a different kind of business. Your own descriptions are probably inconsistent.
- Named only on brand questions: people who already know you find you, but strangers do not. That is the gap AEO and GEO work addresses.
- Competitor appears everywhere: open the sources cited next to them. Often it is a few directories and one or two comparison articles.
Step 5: Turn findings into a to-do list
| Finding | Action |
|---|---|
| Wrong facts appear | Fix the source page, update every profile, then add clear text on your own site stating the correct facts |
| No page answers the question | Write one, answer-first, with real specifics. See answer-first writing |
| Competitor cited from a directory you are not on | Create or claim your listing |
| Cited sources are comparison articles | Reach out to the authors with accurate information and offer a quote or data point |
| Your pages are not cited, but are indexed | Check structure and clarity, and whether the answer is easy to find |
| AI crawlers cannot fetch your site | Review robots.txt and security rules. See AI crawler access |
What this test cannot tell you
- Exact share of voice. Answers vary by run, user and location. Treat results as directional.
- Revenue impact. Pair it with analytics: referral traffic from AI products and branded search trends. See our measurement guide.
- What will happen next. Models and search features change. Re-run monthly.
Automating the audit
Once you have a stable list of 20 to 50 questions, running them by hand every month becomes tedious. A small script can send the same questions to model APIs that support web search, save the full answers and flag changes. This is the kind of monitoring Growmerz builds as part of its AEO and GEO service. One caution: API answers are not identical to what consumers see in the apps, so we use both automated runs and manual spot checks.
A scoring rubric you can reuse
Give each question a score per tool so you can compare months. A simple 0 to 3 scale is enough:
| Score | Meaning |
|---|---|
| 0 | Not mentioned |
| 1 | Mentioned, but wrong, vague or listed far down |
| 2 | Mentioned and broadly correct |
| 3 | Recommended clearly, correct, and one of your pages is cited |
What a filled-in row looks like
These rows are invented for illustration, for a made-up accounting firm. They are not real results.
| Question | Tool | Score | Note |
|---|---|---|---|
| Best accounting firms in Karachi for small businesses | ChatGPT | 0 | Three competitors named, all with directory profiles |
| Is Example Accounting reliable? | Gemini | 1 | Described as "tax consultancy"; the firm offers bookkeeping and payroll too |
| How much does outsourced bookkeeping cost in Pakistan? | Perplexity | 2 | Named, price range correct, source was a third-party article |
From a sheet like this you get three clear actions: build the missing directory profile, rewrite the services description so it names all services, and publish a pricing page the tools can cite.
Writing questions that give clean results
- Write them as a customer would speak, including city and use case.
- Ask one thing per question. "Best bookkeeper in Karachi and how much do they charge" mixes two.
- Do not include your brand name, except in brand-check questions.
- Keep the wording identical month to month.
How long it takes and how to schedule it
| Stage | Time |
|---|---|
| Writing 20 questions | 10 minutes |
| Running them in three tools | 20 to 30 minutes |
| Scoring and noting sources | 15 minutes |
| Deciding three actions | 10 minutes |
Put a recurring reminder in your calendar for the first Monday of each month, and keep the sheet in a shared folder so the history is not lost.
Related guides
- How to Get Your Business Recommended by ChatGPT (What Actually Works)
- How to Measure AI Search Visibility: AEO and GEO Metrics That Matter
- Entity SEO for AI: How Search and AI Decide Who Your Business Is
- What Are AEO and GEO? How to Get Your Business Cited by AI Assistants
- Service: AEO and GEO: Get Found by AI Search
Frequently asked questions
How often should I run an AI visibility audit?
Monthly is enough for most businesses. Run an extra one after major website changes or a new round of reviews and press.
Why do I get different answers each time?
These systems are probabilistic and often use live search. Wording, time, location and account settings all change the answer. That is why we record patterns across many questions.
Is there a free tool that does this?
Some SEO platforms now offer AI visibility tracking, mostly paid. A spreadsheet and 30 minutes works well for a first audit.
What if an AI assistant says something false about my business?
Find the source it cites and correct it if it is wrong. Publish the correct facts clearly on your own site. Most assistants also offer a feedback option on answers. Keep screenshots as a record.
Do I need to audit every AI tool?
Start with the ones your customers use. In most markets that means ChatGPT, Google and one or two others.
Can I trust one tool's answer as representative?
No. Each product retrieves and ranks differently, which is why the audit asks the same questions in several tools.
Sources and further reading
Primary documentation and research behind this guide.
- Google Search Central: Creating helpful, reliable, people-first content
- OpenAI: Overview of OpenAI crawlers
- Perplexity: Bots and crawlers guide
We run this audit for clients every month and send the results as a plain report. If you would rather not do the spreadsheet yourself, book a free AI audit or read about our AEO and GEO service.