How to report AI visibility to clients and leadership

By Miles Burke, founder of CitableReport. Published 18 October 2026.

To report AI visibility, answer four questions every month, in this order: can AI assistants reach and read the site, are we named and linked in their answers, are people visiting from them, and what did we do and what comes next. One number for each, the change since last month, and where the number came from.

Clients and leadership need the same report at different depths. This guide covers which measures to use, where each one comes from, how to be honest about figures that move from run to run, and how to cut the report down for a board. There is a monthly report template to download.

Why is AI visibility hard to report?

It is hard to report because there is no single tool that measures it, and no ranking to point at. A search report can lead with positions and clicks from one console. AI visibility is spread across several places, some of which only count part of the picture.

The answers also change. SparkToro found that ChatGPT and Google's AI almost never gave the same list of brands twice for the same question, even though the same brands tended to keep turning up. A report that says "we rank third in ChatGPT" is reporting a coin toss.

So the job is less about finding the perfect number and more about choosing a few honest ones, measuring them the same way every month, and showing how they connect to things the client already cares about.

What should an AI visibility report show?

A useful monthly report has four parts, each answering one question:

• Can AI reach us? Whether AI crawlers can get into the site and read its pages, shown as an audit score and a list of anything blocking them.
• Are we mentioned? How often AI assistants name the brand, and link to it, when asked the questions customers ask.
• Are we visited? Visits from AI assistants, and what those visitors went on to do.
• What did we do, and what is next? The work done this month and the plan for the next one.

The order matters. If the site blocks AI crawlers, nothing further down will move, so access comes first. And the work done goes last, after the results, so the reader judges the work by what it produced rather than by how busy it sounds.

Which KPIs belong in an AI visibility report?

Keep it to five or six measures. The ones that hold up best are:

• Site score. A repeatable audit of crawler access, page structure, facts and identity. It is the one measure fully in your control, so it shows whether the work was done.
• Share of mentions. Of all the answers collected, the share that name the brand. Our guide to tracking brand mentions in AI sets out how to collect them.
• Citation rate. The share of answers that link to the site, not just name it. Links send visits; names alone usually do not.
• Share of mentions against competitors. The same figure for two or three rivals, so the client sees whether they are gaining ground or losing it.
• Visits from AI assistants, and the enquiries, sign-ups or sales they led to.
• Accuracy. Whether what the assistants say about the business, such as prices, locations and services, is right.

Leave out anything you cannot measure the same way next month. A figure that changes because the method changed is worse than no figure at all.

Where does each number come from?

Each measure has its own source, and each source has limits worth stating in the report.

Google Analytics now has an AI Assistant channel in its default channel group, for visits from assistants such as ChatGPT, Gemini and Copilot. It needs no setup. Google's own AI Overviews and AI Mode are not in it; those visits sit under Organic Search.

Some AI visits arrive with no referrer at all and land in Direct, so treat the channel as a floor rather than a full count. ChatGPT also adds utm_source=chatgpt.com to the links it gives, which helps it to be counted.

Google Search Console includes clicks and impressions from AI Overviews and AI Mode, but counts them within ordinary web search. You cannot split them out, so report the pages that answer your tracked questions rather than claiming an AI Overviews figure.

Bing Webmaster Tools has an AI Performance report, in public preview since February 2026. It shows how often Copilot and Bing's AI summaries cite your pages, and which pages. It does not show clicks, and it covers Microsoft's assistants, not ChatGPT or Gemini.

Your own question tracker gives the share of mentions, citation rate, competitors and accuracy. No platform reports these for you, so they come from asking the same questions every month.

Your site audit gives the site score. Use one with fixed rules, so a change in the score means the site changed, not the auditor's opinion. Our step-by-step audit guide has every check.

How do you report figures that change from run to run?

Report them as shares across many answers, over time, and say so plainly. A share of mentions worked out from 180 answers is steady enough to show a trend. A single answer is not.

Three habits keep the figures honest:

• State the sample. "20 questions, asked three times each on ChatGPT, Gemini and Perplexity, in the first week of the month." One line under the figure is enough.
• Show a three-month trend as well as this month. A rise of two points in one month may be noise; three rises in a row probably is not.
• Keep the questions fixed. Add new ones as a separate set, so last month's figures stay comparable.

Clients take this better than you might expect. Most have seen an AI assistant give two different answers to the same question, and a report that admits it reads as more trustworthy, not less.

What evidence belongs in the report?

Figures persuade the analyst; examples persuade everyone else. Put two or three real answers in each report, as screenshots or quoted text, each with the date, the assistant and the question asked.

Good examples to pick:

• An answer that names and links the client for a category question, especially one that did not last month.
• An answer that names a competitor instead, with the sources it cited. That shows where the work should go next.
• An answer that gets a fact wrong, and what is being done to correct it at its source.

Keep the full record, every answer collected, in a spreadsheet the client can open. They will rarely look, but knowing they can is part of what makes the summary believable.

How should you report AI visibility to leadership or a board?

Cut it down to one slide or one paragraph. Leadership wants to know three things: is it working, what is it worth, and is there a decision to make.

A good board summary might read: "AI assistants now name us in 15% of answers to our customers' questions, up from 9% in June; our main competitor is at 22%. Visits from AI assistants have grown to 412 a month and led to 9 enquiries. We recommend continuing, with a focus on review sites."

Connect the figures to the business in the order shown below, from the work to the money. Leaders are used to marketing measures that only influence revenue, so be clear about which steps you control and which you can only influence.

How often should you report?

Monthly suits clients: it is often enough to show progress and keeps the question tracking manageable. Quarterly suits a board, with the monthly figures available underneath for anyone who asks.

Re-run the site audit after each batch of fixes, not just once a month, so the report can show that the work landed. A score that rises the week after a fix is the clearest evidence an agency has.

What mistakes should you avoid?

• Reporting positions. There are no stable rankings in AI answers. Report how often, not where.
• Mixing up names and links. Being named is good for the brand; being linked is what sends visits. Report both, separately.
• Claiming AI Overviews traffic from Search Console. It is not split out, so any precise figure is a guess.
• Changing the method without saying so. New questions, new assistants or a new tool all break the trend. Note the change and start a new baseline.
• Leaving out what went wrong. A month where the share dropped, explained, builds more trust than a report that only ever goes up.

Is there a template for an AI visibility report?

Yes. Our monthly AI visibility report template (CSV, opens in Excel, Numbers or Google Sheets) has a row for each measure in this guide, with where it comes from, this month, last month and a column for notes. Fill it in each month and the trend builds itself.

For the site score, CitableReport audits a site against 40 fixed checks and gives a score for each of five layers, with two free re-checks after the fixes are made, so the before and after can go straight into the report.

For the mention figures, the tracking guide has its own template that feeds the second part of this one.