An AI visibility report is the monthly document that shows whether and how your brand appears in answers from ChatGPT, Gemini, Perplexity and Google's AI features across a fixed list of questions, alongside AI impressions from Search Console, citations from Bing and visits from AI assistants in GA4. The template we use has 9 rows in 3 blocks (presence, sources, results) plus a mandatory method note; on 27 October 2026, none of the top 10 Google Italy results for "report visibilità ai" offered a template.
So here is the short answer. A useful AI visibility report has few rows, always the same ones, and states how many answers each number is based on.
With Google, reporting was easy. Take Sunscape, which develops solar plants across Italy: from March 2026, in six months, organic impressions went from 329 to 5,809 a month (the full story is in how long SEO takes). One row, one source, one rising line.
With AI, no. The data sits in at least four places and none of them, on its own, answers the marketing director's question: "are we on ChatGPT?".
The mistakes we see most often in the reports clients forward to us:
- a single 0-100 score, with no formula and no sample size;
- months compared on different prompts, so the "growth" is just a change of questions;
- ChatGPT traffic presented as "AI visibility", when it only counts people who clicked.
The big picture, engine by engine, is in our AI visibility guide; this article is about the document that lands on the desk every month.
In this article:
- the 9-row report template, with source and how to read each row;
- a test on a real question put to ChatGPT, and how it goes into the report;
- the rows you already have in your accounts: Search Console, Bing Webmaster Tools, GA4;
- what changes for an agency with several clients, including tools with multi-client dashboards;
- 6 steps to build your first report in an afternoon.
Let's get into it.
What should a monthly AI visibility report include? (9 rows in 3 blocks)
A monthly AI visibility report should include three blocks: your brand's presence in the answers (mentions, position, share against competitors, sentiment), the sources AI engines use (your site, third-party sites, citations measured by Google and Bing) and the results on your site (visits and leads from assistants). Each row has one source and one way to read it.
We explain each metric in AI visibility metrics. Here they are in reading order:
| Block | Row | Source | How to read it |
|---|---|---|---|
| Presence | 1. Mention rate on non-branded questions, per engine | Fixed prompt list | % of answers that name you; the row that matters most |
| Presence | 2. Position in the answer | Fixed prompt list | Average position when you appear |
| Presence | 3. Share of voice against 3-5 competitors | Fixed prompt list | Your share of the brand mentions counted |
| Presence | 4. Sentiment and errors found | Answer text | How many answers describe you wrongly or with old data |
| Sources | 5. Citation rate of your domain | Fixed prompt list | % of answers linking to your site |
| Sources | 6. Most cited third-party domains | Fixed prompt list | The sites to work on next month |
| Sources | 7. Impressions in Google's AI features | Search Console, generative AI report | Trend and pages shown in AI Overviews and AI Mode |
| Sources | 8. Citations in Copilot and Bing | Bing Webmaster Tools, AI Performance | Citations, cited pages, citation share |
| Results | 9. Sessions and leads from assistants | GA4, AI Assistant channel | A minimum, not the total |
And where does the visibility score go?
Outside the table, or next to row 1 with the formula written underneath. On 27 October 2026 we calculated six different scores on the same 10 answers, and the same brand ranged from 8.5 to 67 (the test is in AI visibility score). A number that moves eightfold with the formula should not open a report.
The method note: how many answers each number is based on (without it, the report is worthless)
The method note is the box at the bottom of the report that says how many prompts you checked, on which engines, how many times and on which dates. Without it, nobody can tell whether a 10% rise is a change or noise.
The reason is statistical. AI answers change with every request: in our test on 20 ChatGPT and Gemini answers all 20 brand lists were different.
So every percentage has a margin. With a 60% mention rate:
- on 5 answers the margin is 65 points, almost everything;
- on 20 answers it drops to 39 points;
- on 100 answers to 19 points.
The rule is simple: a monthly change smaller than the margin is not growth. Write it down as "stable".
Our method note always has four items: number of prompts, engines and mode (with or without web search), runs per prompt, collection dates. And a fifth when needed: "the prompt list changed this month, the comparison starts again".
One more split we recommend: branded and non-branded questions in two separate rows. Searchable, which makes a monitoring tool, reports on its own sample that the median brand is named in 93% of answers to branded questions and 7% of non-branded ones (Searchable data, market not specified). Blending them inflates row 1.
How to choose the questions for the list is covered in prompt tracking.
The test: a real customer question put to ChatGPT (and how it goes into the report)
On 27 October 2026 we asked ChatGPT, with web search and location set to Italy, a question that really brings visitors to visilay.com from Search Console: "quali sono i migliori strumenti per monitorare la visibilità sugli llm?" (which are the best tools to monitor visibility on LLMs?). It is one answer, so it works as a reading example, not as a measurement.
What came out:
- 7 tools named: Peec AI, Profound, Otterly.AI, Ahrefs Brand Radar, Semrush, Scrunch and Surfer;
- 13 domains linked as sources, none of them Italian;
- 8 of the 13 domains belonged to the makers of the 7 tools named (Surfer with two domains);
- the other 5 were English-language comparisons, such as technologyadvice.com and brand24.com;
- Visilay and ScryRank: zero mentions, zero citations.
So how does it go into the report?
Like this. Row 1: 0 out of 1. Row 5: 0 out of 1. Row 6: the 13 domains, split between vendor sites and third-party comparisons.
Row 6 is the one that leads to action. If the sources are vendor pages and English comparisons, getting into that answer takes clear comparison pages and mentions on the sites ChatGPT already reads.
One detail made us smile: the answer ended by asking whether we wanted to track a single brand or manage several clients as an agency. We get there shortly.
The rows you already have in your accounts (Search Console, Bing and GA4)
Three rows out of nine need no paid tool: AI impressions in Search Console, citations in Bing Webmaster Tools and sessions in the GA4 AI Assistant channel. They are first-party data, but each one sees a different piece.
Search Console: impressions, not questions
The Search Console generative AI report counts how many times links to your site are shown in AI Overviews and AI Mode, by page, country, device and date. It does not show queries and does not separate AI Overviews from AI Mode, and the table stops at 1,000 rows. The latest days are preliminary, shown with a dotted line: close the month a few days late.
What we saw on visilay.com is in how to track Google AI Mode: 53 queries containing "ai mode" in 90 days, 30 on page one, zero clicks.
Bing Webmaster Tools: the only citations counted by the engine itself
Since 10 February 2026 Bing Webmaster Tools has an AI Performance report: total citations in Copilot and Bing's AI summaries, cited pages, and a sample of grounding queries, the phrases the AI uses to retrieve content. Since 16 June 2026 it also shows citation share, the share of citations your site gets for a grounding query.
It is the only row where citations are counted by the engine, not by a tool asking questions. That is why we keep it even when Copilot matters little in a client's sector.
GA4: the AI Assistant channel is a minimum
In the Google Analytics default channel group guide, the AI Assistant channel collects traffic with the "ai-assistant" medium from ChatGPT, Gemini, Copilot and similar, and excludes AI Overviews and AI Mode, which stay in Organic Search.
On visilay.com, from 29 July to 26 October 2026, that channel collected 54 sessions, 9.3% of the total, yet chatgpt.com showed up under four different mediums. The detail is in ChatGPT traffic in Google Analytics 4. In the report we write "at least 54", never "54".
Why an AI visibility report is not measured in clicks (and what to use instead)
The click is the last thing that happens, and in AI answers it rarely does. Pew Research Center tracked the browsing of 900 US adults in March 2025: with an AI summary on the page, 8% of visits led to a click on a traditional result, against 15% without one, and only 1% clicked a link inside the summary (Pew Research Center, 2025, US data).
Now add our own numbers. 30 queries on page one with AI Mode, zero clicks.
A click-based report would have written "zero" exactly where we were present. And a marketing director who reads zero cuts the budget for the content that works.
That is why rows 1-8 measure presence and only row 9 measures clicks. The order is deliberate.
A report that only counts clicks writes zero exactly where you are present.
You have just seen that 30 queries on page one with AI Mode brought zero clicks. We build your AI visibility report from the questions already in your Search Console, separate presence, sources and results, and tell you which numbers are measured and which are estimated.
Book a strategy callAI visibility reports for agencies with several clients (what changes)
For an agency the monthly report is the same 9-row template, repeated for each client with its own prompt list, locked and dated. Three things change: how projects are managed in the tool, how the report reaches the client and the change log.
What should agencies include in a monthly AI visibility report?
The same 9 rows, plus two items:
- the change log: what was done in the month (updated pages, mentions earned, structured data) and on which prompts an effect is expected;
- competitors chosen by the client, not by you: they are the names the client mentions in meetings.
When we prepared the annual strategy proposal for a manufacturing company we work with, AI visibility went in as two separate chapters, AI Overviews and LLM monitoring, right after the SEO performance report. Separate, because Google measures the first and we measure the second.
In the UK, Hempsall Digital describes a similar structure in its AI search reporting: five layers, from Google AI impressions and Bing citations to a fixed panel of 15 to 20 customer questions, AI crawler access and leads from AI tools.
Which AI visibility tools support multi-client dashboards?
Many, but on different terms and mostly from mid-tier plans. Here is what the official pages state, as checked for our article on how to choose an AI visibility tool:
| Tool | Several clients or brands | Looker Studio / API |
|---|---|---|
| Otterly.AI | Unlimited workspaces from the Standard plan | Both from Standard ($189/month) |
| Peec AI | Agency plans, one project per client | Looker from Advanced, API from Pro |
| Rankscale | 2 to 100 brand dashboards | Looker from Pro, API from Pro+ |
| Profound | Agency plan with a workspace per client | API on Enterprise only |
| SE Visible (SE Ranking) | 3 projects on Basic, no white label | CSV and API |
| Semrush AI Visibility Toolkit | 1 project, multi-brand on Enterprise only | CSV export, API on Enterprise only |
| ScryRank (Visilay) | Not in standard plans | Not stated |
Some tools focus on the client-facing report itself. Ceyo, for example, on its white-label reporting page describes slide-style reports with the agency logo and links that expire after 7, 14 or 30 days, with no login for the client.
Which AI visibility platform is best for an agency managing several clients?
It depends on where the report ends up. If clients already get a Looker Studio dashboard, the connector matters; if they get a PDF, the export matters; if there are many, the price per project matters. We lined up the prices in AI visibility tool pricing.
How often to update the report (monthly, with two exceptions)
An AI visibility report is updated once a month. The two exceptions: more frequent checks while you are changing something on the site, and a new baseline when you change the prompt list or an engine changes model.
Monthly because the noise of single answers only evens out across many checks: more often means charts going up and down for no reason.
Our own tool, ScryRank, is the monitoring tool developed by Visilay and works exactly like this: one check a month on ChatGPT, Claude, Gemini and Perplexity, plus AI Overviews on sector keywords, with a dashboard split into Overview, Visibility, Citations, Prompts, Sentiment and AI Overview. Plans start at €39 a month with 50 prompts, then €69 with 100 and €139 with 200 (price list checked on 27 October 2026). The page does not mention white-label reports: for an agency with many clients that is a limit today, and we say so.
How to build your first AI visibility report in an afternoon (6 steps)
- Pick 25-50 questions starting from the long queries you already have in Search Console, keeping branded and non-branded ones apart.
- Fix 3-5 competitors, the ones your sales team actually meets.
- Decide engines and runs: at least ChatGPT and Google's AI features, and write it in the method note.
- Export the three rows from your accounts: Search Console's AI report, Bing's AI Performance, GA4's AI Assistant channel.
- Fill in the 9 rows and write one sentence per row: what changed and why.
- Close with next month's actions, taken from row 6: the domains to work on and the pages to update.
If you already send a monthly SEO report, the 9 rows go in their own section after organic traffic. How to set up the rest is in SEO report: what to include.
What the report tells you about your chances of being cited (the part you can control)
The report does not make a brand citable, but it shows where to work. Rows 6 and 8 are the operational ones: they show which sites AI engines read on your topic and which of your pages Bing uses as a source.
In our test the sources were vendor pages and English comparisons. For a company selling in its home market the lesson is concrete: you need pages with the facts written out (prices, specs, case studies) and presence on the third-party sites engines already cite, from reviews to industry comparisons. More on this in AI brand mentions and, for the comparison with competitors, in share of voice.
Nobody can guarantee a citation. A well-built report does tell you, month by month, whether you are getting closer.
If you want to set it up with us, our AI Monitoring service starts right here; if the report shows you are missing from the sources, the work moves to GEO. How AI describes your brand will be the topic of our next article on AI brand sentiment.
Frequently asked questions
No. Each tool uses its own metrics and formulas, and the same brand can get very different scores on the same answers. That is why a fixed-row template works better, like the 9 rows in 3 blocks in this article, with the formula and source of every number.
No. The Search Console generative AI report shows your pages' impressions in AI Overviews and AI Mode, but not the questions, not your competitors and nothing about ChatGPT, Gemini or Perplexity. It covers one row out of nine.
It depends: for reading, once a month is enough, because single answers vary a lot and a stable figure needs many checks. While you are changing pages or content, check the affected prompts more often, and when you change the prompt list the comparison starts again.
It depends on the margin you accept: 25-50 questions to start, run several times. On 5 answers a 60% mention rate has a 65-point margin; on 100 answers it drops to 19 points.
No. The GA4 AI Assistant channel only counts visits that arrive with a recognised referrer and excludes AI Overviews and AI Mode, which stay in Organic Search. It measures who clicked, not how often you were named.
Yes. The 9-row structure stays the same, while each client has its own prompt list, competitors and change log. The tool must handle separate projects, which many only offer from mid-tier plans.
Summary
Nine rows, three blocks: presence, sources, results.
A method note with prompts, engines, runs and dates.
Branded and non-branded questions in separate rows.
Search Console, Bing and GA4 give you three rows with no extra tool.
Clicks at the bottom, never at the top.
One last thing the test taught us. The question we asked ChatGPT brings visitors to our site from Google, yet we were not in ChatGPT's answer. That is what an AI visibility report is really for: seeing the gap between where Google finds you and where AI finds you, which are often two different places.
Speak soon!
