How do you check if AI recommends your business?
Ask it — the way your customers would, several times, and write down what happens. Pick 10 questions your customers really ask, run each one in a fresh chat on ChatGPT, Gemini and Google, and record whether you're named, whether your website is linked, and who appears instead. Repeat a week later. It's not scientific, but it's an honest starting point.
Which questions should you test?
The ones your customers ask, in their words. Business owners are often too close to their own business and test the wrong things. Pull questions from your front desk, consultation calls and emails. Good examples:
- "Which dentist in Henderson does same-day crowns?"
- "Who's a good estate planning attorney in Las Vegas for small business owners?"
- "What HVAC company near Summerlin can come out on a Sunday?"
Favor specific questions over broad ones. You're far more likely to win "Spanish-speaking pediatric dentist in North Las Vegas" than "best dentist."
What should you record?
For each question and each AI assistant, note:
- Were you named?
- Was your website linked or cited?
- Who was named instead?
- Was the information about you accurate?
- Which sources did the AI cite for your competitors?
Then ask the AI directly: "Why didn't you mention \[your business\]?" and "What sources did you use for \[competitor\]?" The answers often point straight at the gaps — an outdated listing, a missing directory, a competitor's podcast appearance.
Why won't you get the same answer twice?
AI assistants are experimenting constantly, so the same question can produce different answers from one run to the next. That's why a single screenshot proves nothing. You need repeated tests over time to see a pattern.
Should you pay for an AI visibility tracking tool?
Maybe — but choose carefully. Estimates put the number of AI visibility tools on the market somewhere between 50 and 150, and many of them run only one or two searches per question. That's far too thin to measure anything reliably.
Credible "share of voice" measurement means running each question many times across several AI platforms. Running a set of 30 questions roughly 1,500 times per platform per cycle can bring the margin of error down to about one or two percentage points. That burns real AI processing costs, which is why a very cheap tool can't deliver it.
Also know this: unlike Google's Keyword Planner, OpenAI, Google, Anthropic and Perplexity don't publish how often a given question is asked. Any tool claiming to know AI "search volume" is estimating.
What should you track over time?
- Share of voice: how often you're named across your question set
- Questions won: how many questions name you ahead of competitors
- Citation rate: how often your own pages are cited as a source
- Accuracy: whether what AI says about you is correct
- AI-referred leads: add "How did you hear about us?" to your intake forms and look for AI referrals in your website analytics
Frequently asked questions
How often should I test?
Monthly is enough for most businesses doing it by hand.
Should I test in my own logged-in account?
Use a fresh chat, and ideally a logged-out or private session, so your history doesn't skew results.
What if AI says something wrong about my business?
Fix the source. Find where the wrong information lives online and correct it.
Adapted with permission from Question.Marketing. Statistics are reported findings from the studies named, not DigitalOS guarantees.
