What a monthly AI visibility report should include

AI visibility reporting is still new, and it shows. Some reports are a single “AI score” from a tool nobody can explain. Others are thirty slides of screenshots. Neither tells leadership what they want to know: are we showing up where buyers ask, and is it getting better?

This guide sets out what a useful monthly report includes, how to collect each number, a one-page template you can copy, and what to watch for if an agency sends you one.

The one question the report must answer

“Are buyers hearing our name in AI answers more than last month, and what are we doing about the gaps?” Every number in the report should help answer that. If a metric doesn’t, leave it out, however impressive it looks on a slide.

The five sections

Section 1: Mentions

This is the heart of the report, because mentions are what buyers read. In Pew’s research, people clicked a link inside an AI summary on just 1% of visits that showed one. Being named matters even without a click.

• Mention rate: prompts where you’re named ÷ prompts tracked.
• Share of voice: your mentions ÷ all brand mentions across the prompts, shown next to your top two or three competitors.
• Average position: whether you’re named first, in the middle or last in lists.
• Accuracy: share of mentions that describe you correctly (price, audience, features).

Use a fixed set of 20 to 50 buyer prompts, run in the same assistants every month (typically ChatGPT with search, Perplexity and Google AI Mode or AI Overviews). Run each prompt two or three times, because answers vary, and count a mention if you appear in most runs.

Section 2: Citations

• Citation rate: prompts where one of your URLs is a source ÷ prompts tracked.
• Cited pages: which URLs, for which prompts. This tells you which pages AI trusts.
• Third-party sources: the top domains cited across your prompts, and whether they mention you.

The third-party list is often the most useful part of the whole report, because it turns directly into an outreach list. Our guide to citations vs mentions explains the four states a brand can be in for each prompt.

Section 3: Traffic and conversions

Set up a custom “AI assistants” channel in GA4 covering chatgpt.com, perplexity.ai, copilot.microsoft.com, gemini.google.com and claude.ai. Google says custom channel groups can be applied to your reports retroactively, so you get history too. Report sessions, engagement rate and key events, compared with Organic Search.

Add a caveat line every month: Google counts AI Overview and AI Mode clicks inside normal Google search traffic, and some AI visits arrive as Direct, so this is a floor. Our GA4 guide covers setup.

Section 4: Microsoft AI

Bing’s AI Performance report shows total citations across Copilot, Bing’s AI summaries and some partner integrations, average cited pages per day, page-level citations and a sample of grounding queries. It’s free first-party data, so include it: total citations versus last month, the top five cited URLs, and one line on what the grounding queries show.

Section 5: Wins, gaps and actions

Numbers without actions are a status update, not a report. Close with these four parts, kept short enough to read in a minute:

• Wins: prompts where you’re newly mentioned or cited, and what caused it.
• Losses: prompts where you dropped out, and the likely reason.
• Gaps: the most valuable prompts where you’re missing, with the sources cited instead.
• Next month: three to five specific actions, each with an owner.

Setting it up in month one

1. Build the prompt list. Twenty to fifty buyer questions from sales calls, support, reviews and Reddit. Our prompt research guide shows how.
2. Name your competitors. Two or three that buyers compare you with most.
3. Run the baseline. Every prompt, in each assistant, two or three times. Save the answers and sources.
4. Set up GA4’s AI channel and verify your site in Bing Webmaster Tools.
5. Agree the format with whoever reads it: one page, the template below, same day each month.

The baseline is the most important number you’ll ever report. Every later month is judged against it, so take the time to run it carefully.

The one-page template

AI VISIBILITY, [MONTH]   ([n] prompts | ChatGPT, Perplexity, Google AI)

MENTIONS
Mention rate     [x]%   (last month [y]% | baseline [z]%)
Share of voice   Us [x]% | [Rival A] [x]% | [Rival B] [x]%
Accuracy         [x]% of mentions correct  (issue: [old price on X])

CITATIONS
Citation rate    [x]%   (last month [y]%)
Top cited pages  /compare/[a] ([n] prompts), /pricing ([n])
Top 3rd-party    [domain] ([n]) - lists us? [Y/N]

TRAFFIC (GA4 AI channel)
Sessions [n] ([+/-]%) | Engagement [x]% | Key events [n]

MICROSOFT AI (Bing)
Citations [n] ([+/-]%) | Top page: [url] | Grounding theme: [x]

WINS     Now named for "[prompt]" after [action]
GAPS     Missing from "[prompt]"; cited instead: [domains]
NEXT     1. [action] - [owner]  2. ...  3. ...

Who reads it, and what they need

One report can serve all of them if the top three lines answer the leadership question and the rest is detail for the team. Add a one-sentence summary at the very top, such as “Mentioned in 31% of buyer prompts, up from 24%; still missing from the two biggest ‘best tool’ questions.”

A worked example

Say a company tracks 40 prompts. In month three, mention rate is 28%, up from a 15% baseline. Share of voice is 19%, against 34% for the category leader. Citation rate rose from 8% to 18%, mostly because two new comparison pages are now cited. Accuracy is 85%: two answers still quote last year’s pricing from an old review. AI sessions in GA4 rose 40%, with a higher engagement rate than organic search. Copilot citations doubled, led by the integrations page.

The wins section links the comparison-page citations to the pages published in month one. The gaps section lists the three biggest “best tool” prompts where the company is missing, with the five roundups cited instead. Next month’s actions: pitch those five roundups, ask the old review’s author to update pricing, and publish an alternatives page for the category leader. Anyone reading it knows where things stand and what happens next.

Turning the report into decisions

A report earns its time when it changes what the team does next month. Use simple rules to turn the numbers into decisions:

• Cited but not mentioned on a prompt: your page is helping, but others are recommended. Prioritise outreach to the sources that name competitors.
• Mentioned but not cited: buyers hear your name but your site isn’t the source. Write or improve the page that should answer the prompt.
• Neither: check crawler access and indexing for related pages, then create the missing page and start outreach.
• Accuracy below 90%: fix the sources behind wrong descriptions before anything else, because errors repeat across answers.
• Share of voice flat for two months: look at what changed for the leading competitor, usually new roundup placements or reviews.

The quarterly review

Every three months, step back from the monthly numbers. Look at the trend in mention rate and share of voice across the quarter, which actions moved them and which didn’t, and whether the prompt list still reflects what buyers ask. Add a new group of prompts if your product or market has changed, keeping the original set for comparison. This is also the right moment to decide whether to shift effort between pages, outreach and communities.

How to keep it honest

• Freeze the prompt list. Add new prompts as a separate group, so totals stay comparable.
• Show the baseline every month, not just last month.
• Link everything checkable: the prompt sheet, the GA4 report, the Bing export.
• Report drops as clearly as gains. A report that only ever goes up isn’t believable.
• Note method changes. If you add an assistant or change how prompts are run, say so.

What to leave out

Red flags in agency reports

• The prompt list changes every month, usually towards easier prompts.
• Only wins are shown.
• No competitor comparison.
• No link between the numbers and the work done.
• No raw data you can check yourself.

If you’re choosing an agency, ask to see a sample report before signing. Our 12 questions for AI SEO agencies include what a good one should contain.

How long it takes to produce

About half a day a month by hand, which is a small cost next to the decisions it informs. Paid AI visibility tools can automate the prompt runs; if you use one, make sure you can export the underlying answers and sources, not just a score.

Common mistakes

• Running each prompt once. Answers vary; use two or three runs.
• No competitors. Share of voice is what shows progress in a crowded category.
• Mixing assistants without labels. Report each separately, then combined.
• Ignoring accuracy. Wrong descriptions cost deals.
• No baseline. Without one, nobody can tell if the work is working or if the numbers just drifted.