SEO client reports in 2026: what to include (AI visibility too) and how to prove results

A good SEO client report answers three questions: what changed, why, and what happens next. In 2026 it should also cover AI visibility, alongside organic traffic and rankings: whether the site is citable, whether AI assistants can read it, and whether they recommend the brand over its competitors.

To check it on your own site: For agencies

What is an SEO client report actually for?

An SEO client report exists so that the client can make a decision. It is not there to prove how many hours you logged. If, after reading it, the client cannot say what happened this month and what they need to sign off for next month, the report has failed, however many charts it contains.

Most reports in circulation are tool exports stapled together: a Search Console export, a GA4 export, a list of technical errors and an overall score out of 100. Each piece can be accurate while the whole says nothing. The client sees the score has gone up four points and cannot tell whether that means more sales, more visits or a change in how the tool calculates its score.

Write for two different readers:

• The decision-maker (owner, marketing director) reads the executive summary and, at most, next month's plan. They need five lines and one clear request.
• The implementer (the client's developer or content lead) needs the task list with URLs, severity and the fix. That part can be long, but it goes in an appendix.

Separating those two readers is the single most useful design decision in a report. Everything that follows assumes it: a short body for deciding, and appendices for doing.

What should an SEO client report include in 2026?

A complete SEO client report in 2026 has seven sections: an executive summary, what changed since the last period, traffic and conversions, AI visibility, competitors, the plan for next period, and the data that is missing. The order matters. What the client needs to decide comes first; what they need to trust the numbers comes last.

A five-line executive summary

The executive summary is the only part you can be sure gets read. Five lines are enough if each one does a job:

1. What happened to organic traffic and conversions against a comparable period.
2. The most likely cause, with the figure that supports it.
3. What was fixed from the previous plan.
4. The biggest risk or opportunity you found.
5. What you need from the client this month (a sign-off, access, some copy).

If a line will not fit in one sentence, you do not yet know what you want to say. Write the summary last, once everything else is settled.

Fixed, new and still outstanding

This section proves your work without a single adjective. It is three lists: what the previous audit flagged as failing and now passes, what used to pass and now fails, and what has not changed. That is far more persuasive than "the score went from 58 to 64", because a rising score does not say what was done.

The three lists only hold up if both audits are judged by the same criteria. More on that below, under proving results.

Traffic and conversions (Search Console and GA4)

Performance data comes from two Google tools. The Google Search Console Performance report gives four metrics: clicks, impressions, CTR (clicks divided by impressions) and the average position of the site's topmost result.1 GA4 adds what happens after the click: sessions, conversions and revenue by channel.

Always compare like with like. Month on month is good for spotting a technical problem. Year on year is what you need to talk about trends for any seasonal business: a garden centre's March against its February proves nothing.

AI visibility: access, citability and mentions

This is the new section, and the one that causes most confusion. Split it into three questions, because each one is measured differently:

• Can AI assistants read the site? This is about access: robots.txt, indexing and whether the content is in the served HTML. Search crawlers and training crawlers are different things. OpenAI documents that OAI-SearchBot is the crawler used to surface sites in ChatGPT search, and that sites which opt out of it will not be shown in ChatGPT search answers; GPTBot is a separate training crawler, and each is controlled independently.2
• Is the page citable? A direct answer up front, clear structure, sourced figures, authorship and dates. You audit this on the page itself; our AI visibility audit checklist walks through it block by block.
• Do assistants recommend the brand? This is the outcome: whether ChatGPT, Gemini or AI Overviews name or link to the client for the questions a buyer would ask, and who they cite instead.

The first two are causes, and they can be fixed. The third is an effect, and it is volatile. SISTRIX analysed 82,619 prompts over 17 weeks across six countries, including the UK, and found that ChatGPT replaces between 42% and 74% of the domains it cites every week, depending on the country (60% in the UK, 74% in Germany), and Google's AI Mode between 54% and 59%.3 With that much churn, a one-off screenshot of "ChatGPT recommends us" is not a result. It is a sample of one.

Competitors

Three real competitors are enough: the sites that rank ahead of the client for the queries that matter, not the ones the client thinks are competitors. The useful comparison is not a league table of scores but two lists: what they do that your client does not (gaps) and what your client does that they do not (advantages). Something all three competitors do is probably standard for the sector; something only one does may be chance. How to choose them and read the comparison is covered in our guide to SEO competitor analysis.

The plan for next period

The plan turns the diagnosis into work. Each line says what will be done, on which URL, by whom (your team or the client's) and why it sits where it does in the order. Blockers first (a page that cannot be indexed does not compete), then serious errors, and within each category, whatever has the most headroom.

Keep the plan to what will actually be done in the period. A 60-item plan reads as a list of accusations.

What data is missing, and why

Almost nobody includes this section, and it builds more trust than any other. A missing data point is neither good news nor bad: it is a gap, and it is worth saying where it comes from. A tool may have failed that day, the site may not have responded or may have blocked crawling, or there may not be enough data yet (for example, too little traffic for real-user metrics in PageSpeed Insights).

One line is enough. Saying so protects the client from decisions based on data that does not exist, and protects you from signing off "all clear" on something you never checked.

What do Search Console and GA4 tell you about AI, and what do they miss?

Search Console and GA4 now separate out some AI-related traffic, but each measures something different and neither captures mentions that never get a click. Knowing exactly what each report counts avoids the two most common traps: calling something "AI traffic" when it is not, and blaming a drop in clicks on lost rankings.

Three practical consequences for your report:

1. Google announced the Search Console generative AI report on 3 June 20267 and, according to Search Console Help, it has been available to all sites worldwide since 31 August 2026. If the client cannot see it, the same help page points to two reasons: the site has not had enough impressions in those features, or it has been excluded from them.5
2. GA4's AI Assistant channel does not include AI Overviews. If you total that channel and label it "AI traffic", you are leaving out everything that arrives via Google. Say so in the report, in those words.
3. Zero-click mentions do not show up in either tool. If you want to report on them, you need a separate method (a fixed set of questions put to the assistants with web search switched on, repeated over time) and a clear warning about volatility.

Why fewer clicks does not always mean lower rankings

When an AI summary appears, people click fewer results. Pew Research Center analysed 68,879 Google searches made by 900 US adults in March 2025: with an AI summary, users clicked a traditional result in 8% of visits; without one, in 15%. Only 1% of visits included a click on a link inside the summary itself.8

Ahrefs reached a similar conclusion by a different route. Using aggregated Search Console data for 300,000 keywords, it found that when an AI Overview is present, the average desktop CTR of the top result was 58% lower in December 2025.9 Its earlier study, using March 2025 data, put the figure at 34.5%.10 Ahrefs presents this as a correlation rather than a proven cause, and that is how it should be reported.

For your report, the reading is specific. If impressions hold, average position does not worsen and clicks fall, the most likely explanation is not lost rankings. Show all three metrics together, with the generative AI report beside them. The client then sees that the problem is the format of the results page, not your work, and that the response is to make the page citable within that summary. Our guide to AI search optimization covers what you can and cannot influence.

How do you prove results without inflating the numbers?

You prove an SEO result by comparing two audits run with the same criteria, and by leaving out anything that was missing from either of them. If the tool, the weights or the list of checks changes between one month and the next, the score difference mixes your work with the rule change, and you cannot pull them apart in front of the client.

Four rules make a report defensible:

• The same yardstick on both sides. If the scoring model changes between two audits, recalculate the earlier one with today's model before comparing. If you cannot, say so.
• Something that was not visible before is not a fix. If last month the sitemap could not be downloaded and this month it can and is fine, you have fixed nothing: it is simply visible now. Counting it as an improvement is inflation.
• A missing data point is not a pass. If a check could not be run, do not count it in your favour in either period.
• Correlation is not causation. If traffic rises in the month you fixed the meta descriptions, that may be why, or it may be seasonality. Write "coincides with" when you cannot isolate it.

These rules also protect the client from you. A report that only ever goes up loses credibility; one that explains why a month went down, and what was done about it, wins the next renewal.

When the score drops even though you fixed things

It happens more often than you would think. Fix three minor errors while a serious one appears in the same period (a deployment that breaks the canonical tag, a botched redirect) and the score falls. The fixed, new and outstanding lists settle the argument: the work is done, and the new problem has a name, a date and a URL.

That is why that section comes before the overall score in the template, not after it.

How often should you send an SEO report: monthly, quarterly or by milestone?

For most clients, what works best is a short monthly report plus a longer quarterly review. The monthly report catches technical problems and accounts for the plan; the quarterly review looks at trends, compares year on year and resets priorities.

AI visibility sits better in the quarterly review than the monthly one. Given the weekly churn in sources that SISTRIX describes, a mentions figure taken once a month says little on its own.3 On-page causes, however, can be checked monthly: if someone blocks OAI-SearchBot in robots.txt, you want to know that week.

Between reports, continuous monitoring avoids surprises: a scheduled re-audit of the client's site and an alert when the score drops tell you something has broken before the client notices, and the report is left to explain what happened and what was done. How to set it up, and how often, is covered in our guide to SEO monitoring.

What changes when the report carries your agency's brand?

A white-label report is one produced with a tool and delivered with the agency's logo, colour and contact details, with no trace of the vendor. To the client it is your report, and that has a consequence: you are putting your name to every figure in it.

That changes which tool suits you. If the tool gives figures the client cannot check, or mixes lab and field data without saying so, the error carries your logo. Before putting your brand on someone else's report, check three things:

• Each finding says what is wrong, where and why it matters, in language you can show the client.
• Before and after is calculated with the same criteria on both sides.
• The speed data matches PageSpeed Insights, which the client can open for themselves.

What else to ask of a report that carries your brand is covered in our guide to white-label SEO reports.

Declaration of interest: this blog belongs to SmoothSeen, a web audit tool that measures visibility in search engines and AI assistants and delivers reports under the agency's brand. From the Professional plan, the report comes out as four PDFs and a share link: web ranking, AI ranking, competitor comparison and action plan, plus a link the client opens without an account and that you can revoke. On the Agency and Enterprise plans, white label puts your agency's name, logo, main colour, website and email on the PDFs, the link and the monitoring alert email; any field you leave empty falls back to SmoothSeen's.

For the changes section, the before-and-after view compares against any of the client's 12 previous audits and separates fixed, new and outstanding items. The competitor comparison runs up to three competitors through the same checks as the client's site. Monitoring re-audits each client monthly on Professional, weekly on Agency and daily on Enterprise, and sends an email if a score drops. And on paid plans you can track whether ChatGPT and Gemini name the client for the questions you set.

What SmoothSeen does not do matters just as much when you build the report: it does not connect to Search Console or GA4, it does not track keyword rankings and it does not measure mentions in other assistants. The traffic and conversions sections of the template come from those two Google tools, not from SmoothSeen. Plans are on pricing.

SEO client report template: sections, sources and cadence

This is the structure we suggest, with the data source for each section and how often it is updated. Use it as the report's table of contents: if a section has no source, or nothing new this period, say so in one line rather than padding it.

For section 4b, the guide on how to get cited by AI goes through each point. If the client has a physical location, add the Google Business Profile and the consistency of details between the site and directories to section 4; that is covered in our guide to local SEO for AI search.

A word of caution on section 4a: do not present llms.txt as a visibility win. Google says you do not need to create new machine-readable files, AI text files or special markup to appear in its AI features.4 If the client has one, fine, but it is not something to showcase; the evidence is in our article on llms.txt.

Before-and-after table (template, no data)

This table is a template: the text in angle quotes is a label to fill in, not real data. Use it in section 2 or in a milestone report.

Two rules when filling it in. If a figure does not exist for one of the periods, write "not available" and leave the change column empty. And do not quote a percentage change on a tiny base: going from 2 visits to 4 is a 100% increase that means nothing.

Frequently asked questions

What should an SEO report for a client include?

Seven sections: a five-line executive summary, what was fixed and what is new since the last period, traffic and conversions from Search Console and GA4, AI visibility (access, citability and mentions), competitors, next period's plan and the data that is missing. The technical detail goes in an appendix for whoever implements it.

How often should you send a client an SEO report?

The usual pattern is a short monthly report, which accounts for the plan and catches technical breakages, plus a fuller quarterly review with year-on-year comparison, competitors and AI visibility. Migrations and redesigns deserve their own before-and-after report on top of the regular ones, so that their effect can be separated from everything else.

How do I explain AI visibility to a client?

Split it into three questions: can assistants read the site, is the page easy to cite, and do they actually recommend the brand. The first two depend on the site and can be fixed. The third is a volatile outcome: according to SISTRIX, ChatGPT replaces between 42% and 74% of the domains it cites every week, depending on the country, so report it as a trend.

Why are clicks falling when rankings are stable?

Because Google's AI summaries answer the query on the results page itself. Pew Research found that, with an AI summary, users clicked a traditional result in 8% of visits, against 15% without one. Show impressions, average position and clicks together, with the Search Console generative AI report alongside, so the client can see where the clicks went.

What should I do when a metric could not be measured?

Say so in the report in one line, with the cause: the tool failed, the site blocked crawling, or there is not enough data yet. Do not count it for or against, and leave it out of the trend. If it can be measured next month and turns out fine, do not present it as a fix: it has simply become visible.

What to do next

Take the last report you sent and check whether it has all seven sections, especially the one on missing data. Then audit that client's site with the AI visibility audit checklist and use what you find to fill in section 4 of the next report.