AI Visibility: A 5-Day Check for Local and Service Firms
AI visibility is how often, where and how accurately ChatGPT, Perplexity, Gemini, Claude, Google AI Overviews and AI Mode name your business when someone asks who to hire. The plan below gets you a first read on yours this week. In BrightLocal's Local Consumer Review Survey 2026, use of ChatGPT and other generative AI tools for local recommendations rose from 6% the year before to 45%. Nothing alerts you when an engine names a competitor instead of you or lists a service you dropped. You only find out by looking.
Key points
- The IAB's August 2026 guidelines split AI visibility into four parts: whether you appear, where, how accurately you're described, and whether the answer drives action.
- Wrong facts come in two kinds: one repeats a bad source you can correct, the other has no source, so the fix differs.
- Log the engine, settings, run number and full answer every time, so a recheck compares like with like.
What is AI visibility?
AI visibility is your business's presence inside AI-generated answers: whether an engine names you, where you land in its answer, how accurately it describes you, and whether it actually recommends you. There's no fixed list to look up. The same question can return a different answer each time you ask, so you have to sample it.
In August 2026 the Interactive Advertising Bureau published Measuring Visibility in the AI Era. Its four P's become four things to check in every answer about your firm:
| IAB part | The IAB's question | What you check |
|---|---|---|
| Presence | Does the brand appear in an AI response? | Are you named? Is your site linked? |
| Prominence | Where and how prominently? | First pick, or fourth in a list? |
| Portrayal | In what context, and with what accuracy? | Right services, towns, specialties and category? |
| Persuasion | Does AI visibility drive action? | "The best option for X because of Y," or just "a good option"? |
How do you check your AI visibility this week?
By Friday you'll have a log of every answer, each wrong fact sorted by its cause, and a short fix list.
- Day 1: write eight questions. Cover the four intent types the IAB lists: informational, comparison, recommendation and transactional.
- Day 2: ask each engine each question twice, in a fresh chat every time.
- Day 3: log the four P's for every answer.
- Day 4: trace every wrong fact to the page it came from, if there is one.
- Day 5: read Google's own report, then rank your fixes.
Day 1: what should your questions look like?
Hypothetical example for a small accounting firm, two per intent type, worded the way customers type:
- Informational: "What does a fractional CFO do for a small business?" and "Do I need a CPA for my LLC's taxes?"
- Comparison: "CPA vs. bookkeeper for a small restaurant" and "Local CPA firm vs. online tax service"
- Recommendation: "Best CPA near me for a restaurant owner" and "Good accountant for a dental practice near me"
- Transactional: "CPA near me taking new small-business clients" and "Book a tax planning call with a local CPA"
Day 2: how do you keep the runs comparable?
Run five engines, each as its own row: ChatGPT, Perplexity, Gemini, Google AI Overviews and Google AI Mode. Claude is optional; add it if you have time and log it the same way. Use the same device and location, signed in or out the same way each time. If a Google search shows no AI Overview, log "no AI Overview" and leave that row out of your mention count.
Hypothetical example: eight questions, two runs each, across those five comes to up to 80 answers, or 96 with Claude. That's dozens of answers to run, record and check, so block real time on Days 2 and 3.
Repeats show you how much answers swing. SparkToro's January 2026 study found less than a 1 in 100 chance that ChatGPT or Google's AI, asked the same question 100 times, gives the same list of brands in any two responses.
Day 3: what goes in the log?
Copy these columns into a spreadsheet, one row per answer:
`Date | Engine | Model shown | Signed in? | Location the answer assumed | Run # | Question | Named? | Site linked? | Position | What it said | Full answer (pasted or screenshot link) | Wrong fact? | Source URL shown | Competitors named | "Best" or "an option"? | Fix status`
Keep "Named?" and "Site linked?" apart; they point to different fixes. Keep the full answer beside your summary so you can recheck the exact wording later.
Hypothetical example row: Oct 7 | Perplexity | default | signed out | your town | run 2 | Best CPA near me for a restaurant owner | Yes | No | 3rd of 5 | "offers payroll services" | screenshot link | Yes, payroll was dropped | an old directory listing | Firm A, Firm B | "an option" | listing correction requested
Day 4: what caused each wrong fact?
- A factual inaccuracy is the engine accurately repeating a source that is itself wrong, such as an outdated listing or an old page on your site. Correct the source.
- A hallucination has no basis in any source, or cites a source that doesn't say it. The IAB calls this a platform quality issue, so there's nothing to correct at a source. State the true fact plainly on your site and Business Profile, then recheck next month.
- Source not found. If the answer shows no sources, or you can't find one that says it, log exactly that rather than guessing.
Hypothetical example: in the row above, the statement is "offers payroll services" and the source is an old directory listing. Update the listing, make sure your services page agrees, and once the listing shows the change, rerun the same question in Perplexity with the same settings and log a new row.
Day 5: what does Google's report add?
Search Console's generative AI performance report counts impressions, which Google defines as how many times links to your site were shown to a user in a generative AI feature on Google Search, across AI Overviews and AI Mode. Open the Pages tab, which groups data by the final URL a generative AI feature linked, and set the date range to cover the months before your check. If an old page with outdated details is getting impressions, fix it first, since Google's AI features are linking it. If your main service page is missing, move it up your list. If the report doesn't appear, Google says that usually happens when a site hasn't received enough impressions in those features.
Then rank your fixes:
- Wrong facts with a traceable source go first. They already show prospects bad information about you.
- Named, but only as "an option": spell out your specialties, the clients you serve and your service area in plain text on your site.
- Competitors named and you aren't: compare what their sites and listings state plainly with yours. This guide to getting recommended by AI as a local business covers crawler access, service pages and listings; to hire help, use this checklist for vetting AI SEO companies.
How much can a do-it-yourself check tell you?
Enough to find problems, not enough to grade yourself. The IAB says directional data is not sufficient for budget decisions. It treats fewer than 50 queries per measurement program as exploratory, below even the directional tier. Read your log for wrong facts, missing services and the competitors engines pair you with.
If you'd rather hand off the typing, Vector Index's AI visibility audit runs a bigger version of this check: 25 questions picked from your business, asked four times each on ChatGPT, Gemini, Claude, Perplexity and Google AI Overviews, with every answer stored. Its 25 questions also fall below the IAB's 50-query floor, so it's exploratory too. Its value is the documented record and what comes from it: three change specifications in writing, instead of a spreadsheet to interpret. It doesn't make the changes or guarantee citations. Open the audit page and review its scope and written deliverables before you book.
FAQ
Is AI visibility the same as SEO?
They overlap. Google says there are no additional requirements to appear in AI Overviews or AI Mode, beyond a page being indexed and eligible to show in Google Search with a snippet. ChatGPT, Perplexity, Gemini and Claude need their own checks.
What is a good AI visibility score?
There's no shared benchmark yet. The IAB says there is no common definition of a "mention," and that a brand may get conflicting share-of-voice figures from two different tools. Compare yourself with your own past results on the same questions, and with the competitors that keep showing up beside you.
How often should you check your AI visibility?
Monthly, with the same questions, engines and settings, and again after you change your site or listings. The IAB warns that platform updates can shift visibility metrics overnight, independent of anything you did, so log the date and model of every AI visibility check.