How to write an AI answers report without promising results
What an AI search visibility report from a one-person AI agency should show: fixed questions, dated answers, what changed, and wording that promises nothing.
A physiotherapy clinic owner types "best physio near me" into ChatGPT on a Sunday evening. Three clinics come up. Hers isn't one of them, though she's two streets from one that is. On Monday she emails you: "Can you fix this?"
You can help, and it's a fair service for a one-person AI agency to sell. You can't make a promise about what an AI assistant will say next week, and the report you send her has to make that clear without sounding like an excuse.
An AI answers report records what named AI assistants said when asked a fixed set of customer questions, on a stated date, in a stated way, and compares it with last month. It shows what was said about the business, whether the facts were right, and what you changed. It reports what you observed, never a ranking or a promise.
Why do local businesses want this report?
Because their customers are asking AI assistants for recommendations. BrightLocal's Local Consumer Review Survey 2026, published on 11 February, found AI tools like ChatGPT had risen to third place among the sources consumers use for local business recommendations. How consumers use AI to find local businesses goes through what the survey says and what it doesn't.
Owners hear this and worry. A clear monthly report turns the worry into something they can see and act on. It also gives you a calm answer to "can you fix this?": not a promise, but a record of what the assistants say now, what's wrong in it, and what you'll correct first.
Which questions should you ask?
The ones real customers ask, written down word for word. Five to ten is enough.
For the physio clinic, that might be:
- "Best physio near [suburb]"
- "Physio for a sports injury in [suburb]"
- "Physio open on Saturday in [suburb]"
- "Is [clinic name] any good?"
- "What are the opening hours of [clinic name]?"
Mix questions that don't name the business (does it come up at all?) with questions that do (are the facts right?). Ask the owner which questions their customers ask; they'll know better than any keyword tool.
Keep the list the same from month to month. Change a question and you lose the comparison. If a new question matters, add it and mark it as new.
How do you record the answers fairly?
Write down how you asked, so next month is asked the same way:
| Record | Why |
|---|---|
| Date and time | Answers change; the date makes each one a snapshot |
| Which assistant, and where | ChatGPT, Google's AI Overviews, others; on the web or the app |
| Signed in or not | Signed-in answers can draw on your own history |
| The location used | "Near me" depends on where the question came from |
| The full answer, saved | So you can quote it, not remember it |
| Was the business named? | Yes, no, or only when asked by name |
| Were the facts right? | Address, hours, services, phone |
| Sources the answer showed | Which websites or listings it drew on, if it showed any |
Doing it by hand is fine. Check business in ChatGPT and Google's AI answers walks through a simple routine for it.
What can't the report tell you?
Say this in the report, not only on the call:
- Answers vary. Assistants write each answer fresh. The same question can get a different answer an hour later, or for a different person. One month's result is a snapshot, not a position.
- Different tools, different answers. Google's page on AI features notes that AI Overviews and AI Mode may use different models and techniques, so their answers and the links they show can differ. Other assistants differ again.
- Traffic isn't split out. Google says traffic from AI Overviews and AI Mode is counted in the overall search traffic in Search Console's Performance report. You can't tell the client how many visits came from AI answers alone.
- There's no special trick. Google's same page says there are no additional requirements to appear in AI Overviews or AI Mode, and no special markup to add. The basics that help ordinary search help here too.
That last point matters when someone offers the owner a secret method for getting into AI answers. It's the same answer as the one to can anyone guarantee a top Google ranking: nobody can promise the result.
How often should you run the check?
Once a month is enough for most local businesses, on roughly the same day. Answers vary from day to day, so checking every morning mostly measures that variation. A monthly check, asked the same way each time, shows the trend without drowning the owner in noise.
A few habits keep the comparison fair:
- Same questions, same order, same wording. Copy them from last month's report rather than typing from memory.
- Same setup. Signed out, the same location, the same devices. If you change one, note it.
- Save the full answers, not only whether the business was named. Next month's question is often "what did it say last time?"
- Keep a short log of changes you made between checks, such as fixing opening hours on the Google listing. Without it, you can't tell the owner which change may have made a difference, and neither can anyone else.
If an owner asks for a check more often, say what you'd expect to see: more variation, not more information.
How do you word the report so it promises nothing?
Use words that describe what you saw. Avoid words that describe a position.
| Write this | Not this |
|---|---|
| "Named in 3 of 8 answers, up from 1 of 8 in March" | "Ranking improved" |
| "Opening hours were wrong in one answer; we corrected the Google listing" | "Fixed your AI visibility" |
| "Not named when asked about Saturday appointments" | "Losing to competitors" |
| "We can't tell how many visits came from AI answers" | "AI traffic up 20%" |
A short template for the monthly note:
AI answers check: [month]
Asked on [date], signed out, from [suburb], on [assistants].
Questions asked: [number] (same as last month / [n] new)
Named in: [x] of [y] answers (last month: [x] of [y])
Facts wrong: [what, where]
What we changed: [corrections made]
What we can't know: answers vary by day and person, and Google doesn't split out AI traffic.
Next check: [date]
What work can follow from the report?
The report is useful because it leads to work you can do:
- Correct wrong facts at the source. If an answer gives old opening hours, fix the Google listing, the website and any directory showing the old hours.
- Say plainly what the business does. If the clinic treats sports injuries but its site only says "physiotherapy", add a clear page about it.
- Keep reviews coming. Recent, specific reviews help customers decide, and they're one of the things an assistant may draw on.
Write the service down before you sell it: what the client gets each month, what costs extra, and how it ends. A service recipe is a simple format. If you sell marketing more than websites, selling local services as a marketer covers the same careful approach across other services.
Where Volant fits
Volant's AI visibility service records what AI assistants say about a client, with the date and the method, and asks the same questions the next month so you can compare. You report what you observe, not rankings. It sits alongside the other services to sell after the website.