How to Check Your Business's AI Visibility

Quick answer

Write down the questions your customers actually ask, vary them by service and location, ask them across more than one AI assistant, record whether your business appeared and how it was described, note who appeared instead — then repeat the same questions later and compare.

Why one prompt does not tell you anything

A single prompt tests one phrasing, one service, one location, one assistant and one moment. Answers are generated per question, so two similar phrasings can return different businesses on the same afternoon. Conclusions drawn from one prompt are conclusions drawn from noise.

Consistency is what makes a check meaningful: the same questions, asked the same way, more than once.

Step 1: List the questions your customers ask

Start with how people actually speak, not keywords. For a dental practice: “where can I get an emergency dentist near [city] today?”, “who does Invisalign in [city]?”, “is there a family dentist open on Saturday near me?”

Step 2: Add service and location variations

• One question per major service you offer
• Two or three location phrasings: city, neighbourhood, 'near me'
• At least one urgency phrasing: today, tonight, this weekend
• At least one qualifier your ideal customer would use: affordable, family-friendly, specialist

Step 3: Test across more than one assistant

Different assistants use different sources and behave differently. Testing several gives you a picture rather than an anecdote — and shows whether a problem is general or specific to one system. Our walkthrough of how ChatGPT surfaces local businesses explains why those differences exist.

Step 4: Record the results properly

For each question, capture:

• The exact question text
• The assistant and the date
• Whether your business was mentioned
• How it was described, and whether that description is accurate
• Which other businesses were named

A spreadsheet is fine. What matters is that the question text does not drift between checks — otherwise you are comparing two different tests.

Step 5: Note who appears instead of you

Competitor names are the most actionable part of a manual check. If the same three businesses keep appearing for a service you also offer, look at how plainly their site states that service and location compared with yours.

Step 6: Repeat and compare

Re-run the identical question set on a schedule — monthly works for most local businesses. The comparison is the value: a single snapshot cannot tell you whether last month’s website work achieved anything.

Interpreting what you see

Mentioned, described accurately

Good. Keep the underlying information current and keep monitoring.

Mentioned, described incorrectly

Usually a signal that stale or conflicting information exists somewhere public. Track down the outdated source.

Not mentioned, competitors are

A clarity gap for that question. Our troubleshooting guide covers the common causes.

Nobody local is mentioned

Often the question is too broad or the assistant lacks fresh local data. Rephrase closer to how a customer would speak.

The limits of manual testing

• It is slow, and it decays — most owners stop after two rounds
• Phrasings drift between checks, making comparisons unreliable
• Personalisation and session context can colour your own results
• It is hard to see patterns across dozens of answers in a spreadsheet

Automating the process

StayFront runs a consistent set of realistic customer questions on a schedule, records whether your business appeared and how it was described, tracks which competitors appeared instead, and compares results between scans. What we can and cannot observe is spelled out on our methodology page, and you can preview the output in our sample report.

Frequently asked questions