The short answer
A monthly AI visibility report shows the same agreed questions, asked several times on the same assistants, and how often the client was named ("named in X of Y asks") against a baseline.
Add who was named instead, whether the facts were right, the work done and what comes next, plus one plain line on variation.
Why this is hard to report
Search reporting has settled habits: positions, clicks, impressions, traffic. AI visibility does not have those yet.
- Little or no data. Most assistants give businesses little or no data on how often they are mentioned.
- Answers change. The answers themselves change from one ask to the next. The same question can name your client in the morning and leave them out in the afternoon.
- The screenshot trap. That creates a temptation to report the best screenshot and call it a result. It is also how agencies lose trust: when the client asks the assistant themselves and gets a different answer, the report looks wrong.
A better report sets expectations at the start and measures in a way that survives the client checking.
What to measure
Keep it to a short list of measures you can repeat exactly each time.
- The question set. The questions agreed with the client, in their customers' words. This is the foundation; if it changes, comparisons break.
- How often the business is named. Ask each question several times and count. Report it as "named in X of Y asks", not as a position.
- Whether the site is cited. Some assistants link to sources. Note when the client's own pages are the source.
- Who is named instead. Often the most useful line for the client, because it shows who the assistants see as the alternatives.
- Accuracy. When the business is named, is the description right? Wrong services, prices or locations are worth fixing first.
- The work done. Pages improved, answers added, schema markup fixed, profiles tidied, mentions earned. This is what you control, so show it.
How often to report
- Baseline first. Take a baseline before any work starts. Without it, there is nothing honest to compare against.
- Then monthly. Monthly suits most clients: long enough for changes to have a chance of showing, short enough to keep the work visible.
- A fuller review each quarter. Look at the trend, revisit the question set if the client's business has changed, and agree the next priorities.
Avoid weekly AI visibility reports. They mostly show normal variation, and they invite the client to read meaning into it.
Explaining changing answers
Explain variation once, early and plainly, then repeat a short version in every report. Something like this works:
"AI assistants write a new answer every time they are asked, so the same question can give different results. We ask each question several times and report how often you were named. Small changes from one month to the next are normal. We look for a trend over several months."
When a number moves, give the likely reasons rather than claiming credit or blame straight away. The assistant may have been updated, a competitor may have published something new, or the client's own changes may have been picked up. Say which you think it is, and how sure you are.
Wording to use and avoid
| Avoid | Say instead |
|---|---|
| "You rank #1 in ChatGPT" | "You were named in 4 of 10 asks for this question, up from 1 of 10 at the baseline" (example figures) |
| "Guaranteed visibility in AI answers" | "We will check the agreed questions monthly and report what changed" |
| "We got you into ChatGPT" | "After the changes, you were named more often in our test asks" |
| "You are now optimised for all AI" | "We checked these assistants; other assistants may answer differently" |
| A headline percentage, such as "AI visibility tripled" | The counts themselves: a percentage of a small number of asks exaggerates the change |
| "Your AI score is 82, so you will be recommended" | "The tool estimates the page is easier for assistants to use; it does not predict answers" |
Checking someone else's claims
If an agency or tool tells you a business "ranks" in ChatGPT or Google AI Overviews, ask for the method before the result:
- the question set;
- the dates;
- which assistants were checked;
- how many times each question was asked;
- the baseline.
Then ask a few of the same questions yourself, signed out. A single screenshot or a headline percentage is not evidence of a trend. Our guide to checking whether ChatGPT recommends a business shows how to run those asks.
A report template
One page is usually enough. Use the same headings every time so the client knows where to look.
- Summary in three lines. What changed, what you did, what is next.
- Method. The question set, the assistants checked, how many asks per question, the dates.
- Results against the baseline. Named in X of Y asks, per question or per group of questions.
- Who is named instead. Any new or dropped alternatives.
- Accuracy notes. Anything the assistants got wrong about the business.
- Work done this period. Specific pages and changes, with links.
- Next steps. What you will do before the next report.
- Limits. One short paragraph on variation and on what the method cannot tell you.
The checklist
Use it when you set up a new client and before each report goes out.
AI visibility reporting checklist
Tick items as you go. Ticks stay on this page only and are not saved or sent anywhere.
From geo-agency.co.uk/AiVisibilityReporting. Free to copy and adapt for your own client reports.
Doing this with GEO Agency
The GEO Agency tools can take some of the work off your hands:
- Page analysis to find weak pages.
- Answer simulation to see how an assistant might handle a question.
- Competitor comparison.
- PDF reports that are white-label on paid plans.
They run on Google's Gemini models, so treat their results as estimates and say so in your reports. See plans and prices for what each plan includes.
If you would rather hand the checking over, the done-for-you AI visibility audit covers the question set, repeated asks, a ranked fix list and a re-check. Agencies can ask for it to be run white-label for their clients.
New to the subject? Start with GEO vs SEO: what changes when customers ask AI. To take the baseline yourself, see how to check if ChatGPT recommends a business.
Common questions
How often should I report AI visibility to a client?
Monthly works for most clients, with a fuller review every quarter. Weekly reports tend to show noise rather than change, because answers vary from one ask to the next.
Can I tell a client they rank number one in ChatGPT?
It is better not to. AI assistants write a fresh answer each time and there is no fixed ranking to hold. Say what you measured instead, for example how many of your test asks named the business this month compared with the baseline.
What if the numbers go down after we started work?
Say so, and show what else changed: the assistant, the questions, competitors or the wider web. A drop over one month is often within the normal variation. A drop that lasts several months is a reason to review the plan with the client.
Which assistants should the report cover?
The ones the client's customers are most likely to use, agreed at the start and kept the same. Say clearly in each report which assistants were checked by hand and which results come from a tool that estimates them.
What should a monthly AI visibility report include?
Eight short parts: a three-line summary, the method, results against the baseline ("named in X of Y asks"), who was named instead, accuracy notes, the work done, next steps, and a short note on limits. The template on this page sets them out in order.
Can agencies automate AI visibility reporting?
Partly. Tools can run repeated checks, score pages and produce the PDF, which saves most of the time. A person should still agree the questions, read the answers and write the summary, and the report should say which results come from a tool that estimates them. GEO Agency's tools run on Google Gemini, so their results are estimates.
How can I track AI visibility for all my clients in one place?
Keep one fixed question set per client and record every check in the same format, whatever tool you use. On GEO Agency, the Agency plan (£129 a month, VAT included) adds a shared team workspace for several clients, and paid plans export white-label PDF reports.
How do I sanity-check an agency's claim about AI Overviews or ChatGPT visibility?
Ask for the question set, the dates, which assistants were checked, how many times each question was asked, and the baseline. Then ask a few of the same questions yourself, signed out. Be wary of a single screenshot, a "number one in ChatGPT" claim, or a percentage worked out from a handful of asks.
Who provides monthly AI visibility reporting in the UK?
Some agencies add it to their monthly reports, and some tools let you produce it yourself. GEO Agency is a UK company that offers both: tools agencies use to build their own client reports, and a done-for-you check and re-check, quoted before any work starts.
Should AI visibility replace SEO reporting?
No. Keep the search and traffic reporting the client already relies on and add AI visibility as its own section, with its own method and limits explained.