How to Present an AI Visibility Report to a Client: What Each Section Answers
By TriloAI Editorial TeamPublished
An AI visibility report contains a lot of numbers, but a client really wants three answers: does AI mention us, who do we lose to, and what do we do next. Every other section is evidence for those three sentences.
Below we use the GEO by TriloAI report as the example and go section by section. We don't publish a sample report at the moment, so this article describes the sections in words; if you use another tool, ask the same questions of its report.
Eight tabs, each answering one question
| Tab | What it contains | The client question it answers |
|---|---|---|
| Overview | Trend chart, the GEO score and its six dimensions, visibility by platform, share of voice, cited-source statistics, brand authority, sample size and ranges | Overall, can AI see us |
| Strategy | Comparison with the last analysis, an executive summary for decision makers, and an action list for the people doing the work (immediate, this month, this quarter) | What do we do first |
| Competitors | Brand visibility ranking, lost prompts (questions where a competitor was named instead of you), reasons brands were recommended, competitors' main claims | Who do we lose to, and why |
| AI Responses | The original answer from each platform for each question | Where do these numbers come from |
| Gap Analysis | Questions and platforms where the brand is missing, ranked high, medium or low by opportunity | Which kinds of questions leave us out entirely |
| Content | Draft copy for the channels selected during setup | What can we write first |
| AI Sentiment | The share of positive, neutral and negative mentions | How does AI describe us |
| SEO Audit | Technical site checks, whether AI crawlers are blocked, structured data, and the search queries AI actually ran | Is our site blocking AI, and do we have pages for what AI searched |
Lead with three numbers
- Mention rate: the share of answers that mention the brand. It is the most intuitive number and the easiest to verify, and the report shows it with a 95% range.
- Recommendations and rank: being mentioned is not the same as being recommended. The report counts the two separately and also looks at position; the visibility score weights all of these together (how visibility is calculated).
- Lost prompts: the number of questions where AI recommended a similar brand but not the client. Clients feel this one most, and it converts most easily into actions.
Mention rate, visibility and the GEO score are different numbers
- Mention rate: the share of answers that name the brand. The simplest of the three.
- Visibility: each answer is scored on whether the brand is mentioned, whether it is recommended, its position and how complete the description is; negative mentions get no recommendation credit; platforms are then weighted and combined into 0–100%.
- GEO score: a weighted total of six dimensions: citability 25%, brand authority 20%, content quality 20%, technical SEO 15%, AI visibility 10% and platform readiness 10%. It is closer to "is the site and brand ready to be cited by AI"; whether AI actually mentions the brand is only 10% of it.
So when a client asks whether a score is good or bad, first make clear which score you mean. A monthly report works best built around the mention rate and lost prompts, with the GEO score as a reference for the site's readiness.
AI answers vary: explain it with ranges and "needs confirmation"
Ask the same question today and tomorrow and the AI won't answer word for word the same. In an internal test in September 2026, we re-asked the same 14 questions once and the mention rate moved from 31.5% to 41.7%. That is why the report gives more than one number:
- Sample size: the report states how many questions and answers were used; runs with fewer than 50 questions are labeled exploratory.
- 95% range: the mention rate comes with a range, and if the range spans more than one level the result is marked "needs confirmation", with a suggestion to measure again.
- Re-asking: a full analysis re-asks part of the answers (those mentioning the brand first) once more on ChatGPT, Gemini, Perplexity and Claude; if a brand doesn't appear again, that answer's score for it is halved. The step is skipped when an analysis runs short of time.
- Before and after: the report only compares against an earlier run whose questions overlap enough; if too many questions changed, it shows no before-and-after, and the trend chart doesn't connect the two runs.
When you present it, say the range out loud: "this month's mention rate is roughly between X and Y." It holds up better than a single number and is less likely to be contradicted by next month's movement.
Lost prompts: the section that turns into action
A lost prompt is a question where AI recommended a similar brand but not the client. For each one you can see who the AI picked, why it was recommended, and what the client is missing. Pick the three with the most commercial weight (price and comparison questions, for example) and write one line for each: "the competitor was recommended because of this, and we can add that." That is next month's work plan.
Cited sources: which websites the AI relied on
The cited-source statistics list the domains AI cited and how often. This section sets the direction: if AI keeps citing a particular review site or publication, that is where the client needs to appear; if the client's own domain never shows up, AI search isn't finding the client's pages.
Open the actual URLs instead of reading domain names alone. One domain can host very different kinds of pages; under google.com, for instance, there are map listings and technical documentation. Grouping by domain alone can point the work in the wrong direction.
Checklist before you deliver
- Aliases: enter the client's English name, abbreviations and parent-company name. When AI uses another name and no alias was entered, the mention is counted as missing.
- Off-topic questions: every question should be something the client's audience would ask, with nothing from unrelated industries.
- Competitors in the same category: the rivals AI lists may be different kinds of products; social listening tools and AI visibility tools often share one list. Explain the categories separately.
- Common-word brand names: if a brand name is an ordinary English word, discount any search-volume figures, because people searching that word may not be looking for the brand.
- Report language: market and report language are chosen separately, so check the client can read the language before you send it.
- White-label settings: the PDF's company name, logo and brand color are set in account settings; a share link's company name and logo are set when you create the link.
What goes into the monthly report
- Page one: this month's mention rate (with its range), the number of questions where the brand was recommended, and the change from last month (only when the questions match).
- Page two: the three lost prompts most worth fixing, each with one line of cause and response.
- Page three: changes in cited sources, and whether the client's own domain appeared.
- Page four: what was done this month and what is planned next month.
- Appendix: the SEO audit, content drafts and the original answers by platform, for the client's execution team.
If the Strategy tab shows a projected lift based on current gaps, it is a scenario estimate, not a promise; label it as such in a monthly report, or leave it out.
Delivery: PDF and share links
Exported PDFs can carry the agency's company name, logo and brand color (set under white-label settings in account settings), so the header and footer show your brand. You can also create a share link: choose an expiry of 7, 30 or 90 days or none, add your company name and logo, hide the copy, SEO or competitor sections, and revoke it at any time. Share links still live under geotriloai.com, so if the client must not see the tool's name, send the PDF. How we work with agencies is on the agency plan page.
FAQ
What if the client doesn't understand the GEO score?
Don't start with it. Start with the mention rate and lost prompts, and present the GEO score as a reference for site readiness: six dimensions, with whether AI actually mentions the brand making up only 10%.
Can the report carry our agency's brand?
Yes. The PDF's company name, logo and brand color are set in account settings; a share link's company name and logo are set when you create it. You can resell the report to clients, and we don't contact an agency's clients directly (see the agency plan).
How often should we re-run it?
Once a month is enough for most clients; what matters is using the same questions each time. To check a site change, run it again a few weeks later; each full analysis uses 3 credits. For planning the monthly deliverable, see how agencies can offer GEO.
Curious whether ChatGPT recommends your brand right now
Free quick scan (ChatGPT only, about 1–2 minutes) →