GEO report for clients: a monthly template with metrics and margin of error

A GEO report template for agencies to deliver every month: the seven sections, the metrics and the source of each, how to show the margin of error, ready phrases to explain swings and a method appendix, all in copy-ready blocks and without promising AI rankings.
Written with AI, reviewed by humans

In this article

See how your brand shows up in AI answers

Book a demo and find out what to fix first, with a GEO specialist.

Book a demo

Reading time: 10 minutes

A GEO report for clients covers, every month, the brand’s presence in AI answers, the traffic that comes from them and the next actions, always with a margin of error. This monthly GEO report template brings sections, metrics and copy-ready blocks, because the format repeats every month. So the agency doesn’t promise rankings or citations.

In short

  • The GEO report for clients has seven sections, from the executive summary to the method appendix. So each number comes with a source, a comparison and a margin of error.
  • Presence in answers comes from fixed prompts, while traffic, impressions and citations come from GA4, Search Console and Bing. For that reason, the sources never add up.
  • Therefore, with 1,000 answers per month, only a change above 4 to 10 points, confirmed the next month, counts as real.
  • The report commits to method and deliverables, never to rankings or citations, because the chance that ChatGPT or Google’s AI repeats the same list of brands is below 1 in 100.

What should a GEO report for clients include?

A GEO report for clients needs seven sections, from the executive summary to the method appendix. The order starts with what the client decides, that is, the business, and ends with what the agency will do next month. Google, by the way, suggests that clients ask the SEO professional how success is measured. So the report answers that question every month.

SectionWhat it showsSource
Executive summaryResult, likely cause, next stepsAgency text
BusinessLeads, sales and sessions from AIGA4 and CRM
Presence in answersPresence, citation, recommendation, share of voiceFixed prompts
Google and MicrosoftImpressions and citations in AI featuresSearch Console and Bing
ReputationSentiment and errors about the brandBrand-group prompts
ActionsThree to five actions with owner and deadlineAgency plan
Method appendixPrompts, engines, country and changesCollection log

Who reads each section of the report?

The GEO report for clients has two readers, because whoever decides does not read the rest. The decision maker reads only the summary, while the client’s team checks the other sections. So the summary goes on the first page, and the rest serves as proof.

The five copy-ready blocks below form the full GEO report template, but the fields in [BRACKETS] change for each client. For example, if the client only buys SEO, the AI block goes into the SEO report template for clients, which also covers traffic and conversions. Since the list of indicators, including the misleading ones, is in the guide to GEO metrics, this template doesn’t repeat it. If the concept is new to the client, then start with the guide to what GEO is.

GEO report for clients: illustration of a monthly report with charts of brand presence in AI

How do you write the executive summary of a GEO report?

In a monthly GEO report, the executive summary fits in five lines: result, change, likely cause, risk and next step. Write it last, but put it on the first page, because the decision maker reads only that. So each line carries the number, the comparison with the baseline and the margin of error.

For example, if the change stays within the margin, the text says the brand is stable, not that it improved. That way, the client doesn’t demand a rise that the data can’t support. On the other hand, if the change passes the margin, the text asks for confirmation next month instead of celebrating.

Copy-ready block 1: executive summary

EXECUTIVE SUMMARY ([MONTH])
1. Result: the brand appears in [X]% of
   answers (± [M] points), against [Y]%
   at baseline.
2. Change: [rise, drop or stable]. It is
   [within or above] the margin.
3. Likely cause: [agency action or external
   change], without claiming causation.
4. Risk: [error about the brand, rising
   competitor or a lost source].
5. Next month: [action], with [OWNER]
   and a deadline of [DATE].

The example below is illustrative, because the numbers are made up. So it shows, in the same month, a stable reading and a change above the margin.

Filled-in example (made-up numbers)

EXECUTIVE SUMMARY (ILLUSTRATIVE EXAMPLE)
1. Result: the brand appears in 31% of
   answers (± 5 points), against 29%
   at baseline.
2. Change: presence is stable (+2 points,
   within the margin). Site citation went
   from 12% to 25%: above the margin, so it
   counts as a gain only if it repeats
   next month.
3. Likely cause: the pricing page was
   rewritten this month, without claiming
   causation.
4. Risk: one answer got the plan price
   wrong, and the cited source is a directory.
5. Next month: fix the directory and
   run the brand prompts again.

Which metrics go into a GEO report, and where does each come from?

In a monthly GEO report template, presence, citation, recommendation and share of voice go in, plus sessions, impressions and AI citations, because each answers a client question. The first four come from fixed prompts, run every month, since AI changes its answers. On the other hand, the rest come from GA4, Search Console and Bing, and each source counts something different.

MetricHow to calculateSource
PresenceAnswers with the brand ÷ totalFixed prompts
CitationAnswers citing the site ÷ totalFixed prompts
RecommendationAnswers that recommend the brand ÷ totalFixed prompts
Share of voiceBrand mentions ÷ brand and competitorsFixed prompts
AI sessionsSessions from the AI Assistant channelGA4
AI impressionsImpressions in Google’s AI featuresSearch Console
AI citationsCitations in Copilot, Bing and partnersBing Webmaster Tools

Which caveats apply to each source?

So three caveats prevent misreading. First, the GA4 AI Assistant channel excludes Google’s AI Overviews and AI Mode. Those clicks stay in Organic Search, so AI sessions show only part of the effect.

Second, the Search Console generative AI performance report shows impressions, not visits. Finally, Bing’s AI Performance shows citations, but the total doesn’t indicate placement or presentation within an answer. So call this data citations, not traffic.

Reputation, on the other hand, comes from the brand-group prompts. Classify each mention as positive, neutral or negative with a fixed criterion, and note each wrong piece of information, with the source the AI cited. After all, the fix is often outside the site, in a directory or an old article.

The formula for each rate and the collection protocol, by the way, are in the guide to how to measure brand visibility in AI.

Copy-ready block 2: numbers page

PRESENCE IN ANSWERS ([MONTH])
Engine     Presence  Citation  Recom.  Base
ChatGPT    [X]%      [X]%      [X]%    [X]%
Gemini     [X]%      [X]%      [X]%    [X]%
AI Overv.  [X]%      [X]%      [X]%    [X]%
AI Mode    [X]%      [X]%      [X]%    [X]%
Share of voice vs [N] competitors: [X]%
Answers this month: [N]. Margin: ± [M] points.
Margin for each engine: ± [MM] points.

GOOGLE AND MICROSOFT
Sessions, AI Assistant channel: [N]
AI impressions (Search Console): [N]
Bing citations (AI Performance): [N]
These numbers do not add up.

How do you show the margin of error to the client?

In a monthly GEO report, show the margin next to each rate and put the reading rule in the footer, so the client sees it with the number. The margin depends, therefore, on the number of answers.

For example, with 1,000 answers per month, from 50 prompts, 5 runs and 4 engines, it stays between ± 3 and ± 7 points. With 250 answers, however, it rises to ± 6 to ± 14. The ranges, by the way, are our own calculation, at 95% confidence. The rate for each engine, however, uses only 250 of those answers, so its margin is that of the 250 row.

The reading rule is short, but it demands discipline. After all, since both months have error, the difference between them must pass about 1.4 times the margin. So, with 1,000 answers per month, treat only a change above 4 to 10 points, confirmed the next month, as real.

So smaller changes go into the report as stable. In addition, the limit goes into the contract, as the GEO proposal template shows. That way, the GEO report template and the proposal use the same ruler.

Copy-ready block 3: margin of error footer

HOW TO READ THE RATES
Margin of error, at 95% confidence:
- 30 answers in the month: ± 18 to ± 31 points
- 250 answers: ± 6 to ± 14 points
- 1,000 answers: ± 3 to ± 7 points
This month: [N] answers, margin ± [M].
Change below [4 to 10] points = stable.
Above that, it counts only if it repeats
next month.

How do you explain a swing or a drop in a GEO report?

In a GEO report for clients, explain the swing by pattern and margin, because AI answers change with every run. According to a study by SparkToro, the chance that ChatGPT or Google’s AI gives the same list of brands in two responses was below 1 in 100.

The study, for example, used 600 volunteers and 2,961 runs of 12 prompts. However, it involved a researcher from Gumshoe.ai, an AI tracking startup, and the author discloses the possible conflict of interest. Even so, the data helps explain to the client why a swing within the margin needs no cause, while a drop above it calls for investigation.

Use three readings, always in this order, so the text is the same every month:

  • Stable: the change stays within the margin, so the report only records it.
  • Change above the margin: the report shows the number and waits for next month, to call it a gain or a loss.
  • Collection change: the report warns that the comparison with last month lost strength, because prompts, engines or competitors changed.

In a confirmed drop, then list the causes to check: the collection, the monitored competitors, site changes and the third-party sources the AI cited. So each cause gets an owner and a deadline. At renewal, by the way, this history carries weight, as the guide to GEO for agencies explains.

Copy-ready block 4: ready reading phrases

STABLE
"Presence stayed at [X]%, within the margin
of ± [M] points. There was no change."

RISE ABOVE THE MARGIN
"Citation rose [N] points, above the margin.
We will confirm next month before
counting it as a gain."

DROP ABOVE THE MARGIN
"Presence fell [N] points, above the margin.
Causes to check: [LIST]. Action: [ACTION]
by [DATE]."

NO PROMISE
"The report shows method and deliverables.
The agency does not commit to citations."

How do you adapt the template to a client with little data?

With little data, the monthly GEO report changes its role, since instead of measuring monthly change it shows trend and diagnosis. With 30 answers per month, the margin runs from ± 18 to ± 31 points, so almost no monthly change passes it. Then, by our criterion, group the months into a quarterly review.

So use the month for qualitative findings: which domains the AI cites, which errors appear and which competitor shows up in the answers. GA4, Search Console and Bing, however, count events and remain useful. So the GEO report template fits small contracts, as long as it says what the number can’t decide yet.

What goes in the method appendix, and when do you deliver the report?

In the monthly GEO report template, the appendix records everything that changes the answer, except time. It lists engines, country, language, prompt list, runs, competitors and changes since last month. Without it, after all, nobody can redo the math next month. In addition, a competitor swap looks like a change in AI, and only the appendix shows it was the collection.

As for timing, the deadline depends on GA4, because processing takes time. According to the Google Analytics help, processing can take 24 to 48 hours, and the credit for key events can change up to 12 days after they are recorded. So, by our criterion, close the GA4 numbers only after the 12th day of the following month, or warn that the credit by channel may still change.

Search Console, however, uses the Pacific time zone for dates. So the monthly cutoff may differ from GA4, and the report states the period of each source. Finally, by our criterion, review the prompt list and the goals every quarter, not every month.

Copy-ready block 5: method appendix

METHOD APPENDIX ([MONTH])
Engines: [LIST]. Country and language: [US, en].
Prompts: [N], list frozen until [DATE].
Runs: [N] per prompt and per engine.
Answers collected: [N]. Margin: ± [M].
Monitored competitors: [LIST].
Collection: new chat, no personalization,
through the interface [yes or no].
GA4 closed on: [DATE].
Search Console and Bing period: [DATES].
Changes since last month: [LIST].

Frequently asked questions

Can the GEO report and the SEO report be the same document?

They can, since the AI block goes into the monthly SEO report when the client buys both SEO and GEO. It sits next to traffic and conversions, as the SEO report template for clients shows. What you can’t do is add presence in answers to traffic, because each number comes from a different source.

What should the report show when the brand appears in no answers?

Then show the rate of 0%, with the margin of error and the baseline, like any other result. Next, list the causes to check and the month’s actions, each with an owner and a deadline. After all, a recorded zero is the starting point for comparing next month, not a reason to hide the metric.

Can you build a GEO report for clients without a paid tool?

You can, while the volume is small, because manual prompt collection, GA4, Search Console and Bing Webmaster Tools cost nothing. When the volume passes a few hundred answers per month, collection moves to a tool or a script, as the guide to measuring brand visibility shows.

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Main sources

About this articleWritten with AI and reviewed by humans, with verified sources. Found something out of date? Write to the NotHuman team.

Get new NotHuman articles by email

By subscribing, you agree to receive blog articles by email and to the Privacy Policy. Unsubscribe any time.

Keep reading

Demo

Want to know how your brand shows up in AI answers today?

Book a demo and see how your brand shows up in AI answers, with support from a GEO specialist.

Book a demo

Your privacy

We use necessary cookies to make the site work and Google Analytics and Microsoft Clarity statistics cookies to improve our content. You can decline the statistics ones; in the EU, UK and Switzerland they are only turned on after you accept. Cookie Policy.

Cookie preferences

Cookie Policy