AI Search · 16 min read
SEO reporting in 2026: the metrics that matter when AI answers first
SEO reporting for AI search: the GA4 setup for AI traffic, how to measure AI share of voice and citations, and a one-page monthly report for your board.

On this page — 14 sections
- TL;DR
- Table of Contents
- Why are organic clicks falling when visibility isn’t?
- Being cited is the new position one
- How to track AI traffic in GA4
- What referral data misses
- How to measure citation share and AI share of voice
- The one-page monthly SEO report for your board
- What to stop reporting
- Our perspective
- Want reporting your board will trust?
- FAQ
- Sources
- Recommended
SEO reporting in 2026 has to measure whether buyers see and trust your brand inside AI answers, not just whether they click through to your site. That means adding three things to the usual report: how often AI Overviews, ChatGPT, Gemini, Perplexity and Copilot cite or mention you (citation share and AI share of voice), how much traffic AI assistants send you (a dedicated AI channel in GA4), and what that visibility turns into (conversions and pipeline). Raw rankings and total sessions on their own now tell a board very little.
If you’re a marketing director, you’ve probably had the awkward meeting already. Organic clicks are down year on year, the agency report says rankings are fine, and someone on the board asks whether SEO still works. Both things can be true at once, and this article explains why, then shows you how to report it properly: the GA4 setup, the AI metrics worth tracking, and a one-page monthly layout you can put in front of a board.
Need a reporting setup that shows your AI visibility, not just your rankings? Castle builds AI search monitoring and reporting into every AI SEO engagement.
TL;DR
- Clicks are falling where AI answers appear: Ahrefs found a 58% lower click-through rate for the top-ranking page when an AI Overview shows, and Pew Research found people clicked a result in 8% of visits with an AI summary against 15% without.
- Being cited still pays. In Seer Interactive’s 2026 update, brands cited in an AI Overview earned 2.07% organic CTR on informational queries against 0.94% for brands that weren’t.
- Set up a separate AI channel in GA4 with a regex for chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai and the rest, placed above Referral.
- Referral data undercounts AI badly: most AI apps strip the referrer, and zero-click answers never reach your site at all.
- Report citation share and AI share of voice from a fixed prompt panel, alongside AI referrals, conversions and pipeline. Drop raw rankings in isolation.
Table of Contents
- Why are organic clicks falling when visibility isn’t?
- Being cited is the new position one
- How to track AI traffic in GA4
- What referral data misses
- How to measure citation share and AI share of voice
- The one-page monthly SEO report for your board
- What to stop reporting
- Our perspective
- Want reporting your board will trust?
- FAQ
- Sources
Why are organic clicks falling when visibility isn’t?
Clicks are falling because Google now answers a large share of questions on the results page, so fewer people need to visit a website at all. Your pages can be read, summarised and cited by Google’s AI without anyone clicking. The impression still counts. The click doesn’t happen.
The studies agree on the direction, if not the exact size:
- Ahrefs compared 300,000 keywords in December 2023 and December 2025 and found that the presence of an AI Overview correlated with a 58% lower average click-through rate for the top-ranking page. Its April 2025 study had put the drop in clicks at 34.5%, so the gap has widened.
- Pew Research Center tracked the browsing of 900 US adults and found users clicked a traditional result in 8% of visits when an AI summary appeared, against 15% without one. Only 1% of visits involved a click on a link inside the summary, and people ended their browsing session on 26% of pages with an AI summary, against 16% without.
- Seer Interactive analysed 5.47 million queries across 53 brands. In December 2025, organic CTR on queries showing an AI Overview was 1.31%, against 3.16% where none appeared. It recovered to 2.36% by February 2026, still well short of the 3.82% on queries without one (Seer Interactive).
- Advanced Web Ranking, whose CTR study covers the UK as well as the US, says the position-one click loss when an AI Overview appears has been “typically on the order of half”, varying by month, device and intent.
One honest caveat. Most of these datasets are American, and we haven’t seen a UK study of the same size. The habit is clearly here, though: Ofcom reported in April 2026 that 54% of UK adults now use AI tools such as ChatGPT, Copilot or Gemini. Treat the US numbers as the direction of travel, not a forecast for your account.

There’s a second problem, and it’s one of plumbing. Google says traffic from AI Overviews and AI Mode is included in the overall “Web” search type in Search Console, with no separate filter. So when your impressions hold steady and clicks drop, Search Console can’t tell you how much of that came from being cited in an AI answer, or how much from a competitor being cited instead. You need other measures for that, which is what the rest of this article is about.
Being cited is the new position one
If you’re going to appear in an AI Overview anyway, you want to be one of the sources it names. Seer’s 2026 update is the clearest evidence for this. On informational queries across 2025, brands cited in the AI Overview averaged a 2.07% organic CTR. Brands on the same results pages that weren’t cited averaged 0.94%. Seer puts that at about 120% more organic clicks per impression for cited brands. Paid ads showed a citation premium too, of more than four percentage points every month in 2025.
Being cited doesn’t get you back to the old days. The same analysis shows cited brands still trailed queries with no AI Overview by around 38%. But the choice in front of you isn’t “AI Overview or no AI Overview”. Google makes that call. Your choice is whether you’re the brand in the answer or the brand underneath it.
Think of it like a shop window on a busy high street. Plenty of people stop, look and walk on, and you’ll never see them in your till receipts. Some of them come back next week because they remembered the window. A report that only counts people through the door misses most of what the window is doing.
That’s why reporting needs a visibility measure that isn’t clicks. We cover how to measure it below. If you want the groundwork first, our guide to earning LLM citations explains what makes a page quotable.
How to track AI traffic in GA4
Out of the box, GA4 lumps visits from ChatGPT, Perplexity and the rest into Referral, mixed in with directories, partner sites and spam. Pulling them into their own channel takes about ten minutes and gives you a clean trend line from day one.
Step 1: copy the default channel group
In GA4, go to Admin, then Channel groups under Data display. You need Editor access or above. Copy the default channel group and give it a name such as “Channels with AI”. A standard property allows two custom channel groups alongside the default one, with up to 50 channels in each, so use one slot wisely.
Step 2: add an “AI assistants” channel with this regex
Add a new channel called “AI assistants”. Set the condition to Source and choose matches regex, then paste:
.*(chatgpt\.com|chat\.openai\.com|perplexity\.ai|gemini\.google\.com|copilot\.microsoft\.com|copilot\.com|claude\.ai|chat\.deepseek\.com|deepseek\.com|meta\.ai|grok\.com|chat\.mistral\.ai|you\.com|poe\.com).*The .* at each end means the pattern still works if GA4 treats it as a full match, and it catches variants such as www.perplexity.ai. ChatGPT also adds utm_source=chatgpt.com to the links it sends, so those visits arrive with chatgpt.com as the source even when a referrer is missing. Review the list every quarter. New assistants appear, and domains change.
Step 3: move it above Referral
This step is the one people skip. GA4 puts each visit into the first channel whose definition it matches, working down the list in order. If “AI assistants” sits below Referral, Referral claims the visits first and your new channel stays empty. Drag it above Referral and Organic Search, then apply.
Custom channel groups apply to your reports retroactively, so you’ll see historic AI traffic straight away rather than waiting months for a baseline.
Step 4: report landing pages and conversions, not just sessions
Build a simple exploration with the new channel group, landing page and your key events. AI referrals are often small in number and high in intent, so the useful questions are which pages AI assistants send people to and whether those visits convert. Session counts alone will undersell it.
What referral data misses
Referral data gives you the visits AI sends with a label attached. It misses most of the influence, for two reasons.
First, many AI apps don’t pass a referrer. SearchPilot’s May 2026 analysis of how AI traffic shows up in analytics found that links opened from the ChatGPT mobile app arrive as Direct, describing it as “near-100% dark”. The Claude apps behave the same way, and around 91% of Gemini app traffic was dark. Perplexity’s app was the best behaved, with roughly 70% of sessions showing as Referral. So a buyer who researches you in the ChatGPT app on their phone, then taps through to your pricing page, looks like someone who typed your URL from memory.
Second, the most valuable AI moment often involves no visit. A finance director asks Copilot for three UK suppliers of what you sell. You’re named. She books a call two weeks later after searching your brand name. GA4 credits branded organic search or Direct. Nothing in your analytics connects that enquiry to the AI answer that put you on the list.

It’s no surprise the measured numbers look small. Conductor’s 2026 benchmarks, reported by Search Engine Land, put AI referrals at 1.08% of all web traffic across 13,770 US domains, with ChatGPT sending 87.4% of them. If you judged AI search on that figure alone, you’d conclude it doesn’t matter. That’s the reporting trap.
You can’t fix the plumbing, but you can triangulate with a few proxies:
- Branded search trend in Search Console. If AI answers are recommending you, more people search your name.
- Direct traffic to deep pages. Hardly anyone types a long service-page URL from memory. A rise in Direct sessions landing on deep commercial pages is often dark AI traffic.
- A “How did you hear about us?” field on enquiry forms, with ChatGPT, Gemini and “an AI assistant” as options. It’s crude, and it’s the most direct evidence you’ll get.
- Sales call notes. Ask your team to log when a prospect mentions an AI tool. Our comparison of GEO and Google leads covers why this matters for lead quality.
How to measure citation share and AI share of voice
Citation share is how often AI answers link to your site as a source. AI share of voice is how often your brand is mentioned compared with the competitors you track, across the same set of prompts. Ahrefs gives the basic formula as responses that mention your brand, divided by responses that mention any tracked brand, multiplied by 100. Both are measured against a fixed list of questions, which is the part most businesses get wrong.
Build a prompt panel
A prompt panel is a fixed list of the questions your buyers ask AI assistants. Start with 30 to 50. Mix the stages:
- Category questions: “Best commercial insurance brokers for manufacturers in the UK”
- Comparison questions: “[Your brand] vs [competitor] for fleet management”
- Problem questions: “How do I cut energy costs in a cold-storage warehouse?”
- Brand questions: “What does [your brand] do?” and “Is [your brand] any good?”
Keep the panel fixed from month to month. Change the prompts and you’ve changed the measurement, and the trend becomes meaningless. Add new prompts as a separate set if you need to.
Run it with a tool
Doing this by hand across six assistants every month gets tedious fast. Ahrefs Brand Radar tracks mentions, citations, AI share of voice and estimated impressions across AI Overviews, AI Mode, ChatGPT, Perplexity, Gemini and Copilot, with Claude available for custom prompts. Its methodology page explains that prompts run through the public versions of each assistant. Other tools work similarly. Whichever you use, check that you can fix the prompt set, the country and the competitor list.
Keep some manual checks
Tools sample. A person reads. Once a month, run your top ten money prompts by hand in ChatGPT, Gemini and Copilot and note three things: are you named, is what’s said about you accurate, and which pages or third-party sites are cited. Inaccurate descriptions are often a bigger risk than absence, and a tool’s share-of-voice number won’t flag them.

Read the numbers with care
AI answers vary between runs. Ahrefs’ own advice is to trust trends that persist across several reporting periods rather than one-off spikes, and it warns that a blended score can hide a weak platform: a strong ChatGPT number can mask a poor Gemini one. Report each assistant separately where you can, and treat a single month’s movement of a few points as noise.
The one-page monthly SEO report for your board
A board needs to know whether buyers can find you, whether they trust what they find and whether it’s producing revenue. Everything else belongs in an appendix for the marketing team. Here’s the layout we’d use, in this order:
- 1. Visibility: Search Console impressions for non-branded commercial queries, month on month and year on year, with a one-line note on how many of your tracked keywords now show an AI Overview.
- 2. AI citations and share of voice: your share of voice across the prompt panel, split by assistant, against your three main competitors; citation count for your domain; any inaccurate descriptions found and fixed.
- 3. AI referrals: sessions from the “AI assistants” channel, the top five landing pages, and the trend in branded search and Direct to deep pages as the dark-traffic proxy.
- 4. Conversions: enquiries, sales or key events from organic search and AI assistants, shown separately, plus the conversion rate for each.
- 5. Pipeline: qualified leads and pipeline value influenced by organic and AI, including any self-reported “found you via AI” enquiries, taken from the CRM rather than GA4.
- 6. Three actions for next month: what you’ll fix or publish, tied to a gap in the numbers above.
Keep the commentary to two or three sentences per section and lead with the “so what”. For example, “Clicks down 12%, impressions up 4%, share of voice up from 18% to 23% in ChatGPT, enquiries flat” tells a board that visibility is growing while the click model changes underneath it. That’s a very different conversation from “traffic is down”.
If you want a structured way to find the gaps that feed section 6, our AI visibility audit checklist is the companion to this piece.
What to stop reporting
Some staples of the old SEO report now actively mislead. Drop these, or at least stop leading with them:
- Raw rankings in isolation. Position three under an AI Overview with four cited sources is a very different asset from position three on a plain results page. If you report rankings, report them with AI Overview presence and whether you’re cited.
- Total organic sessions as the headline. Sessions fall as AI answers more questions, even when you’re doing well. Lead with commercial outcomes.
- Vanity traffic. Blog posts that rank for broad informational terms and send visitors who never convert. AI now answers most of those questions directly, so the traffic is going anyway.
- Keyword counts. “We rank for 4,200 keywords” says nothing about whether any of them make money.
- Average position across everything. It blends branded, non-branded and junk queries into one number nobody can act on.
None of this means traffic doesn’t matter. It means traffic is now one output among several, and the board deserves to see the others.
Our perspective
We think the industry spent too long reporting the metric that was easiest to pull rather than the one that mattered. Rankings and sessions were convenient. They’re now incomplete, and an SEO report that leans on them will either look like failure when it isn’t or look like success when a competitor is quietly taking your place in the answers.
At Castle, we pull GA4 and Search Console data into Claude through MCP connectors, run a fixed prompt panel across the main assistants and read the answers ourselves before anything goes in a report. Automation does the collection. People do the judging, because an AI tool can tell you that you were mentioned but not whether the description would put a buyer off.
Our position is simple. Report on whether buyers see and trust you in AI answers, not just whether they click. If you’re cited, described accurately and turning up in pipeline, you’re winning, whatever the session count says. Our AI SEO, GEO and AEO work is built around that.
Want reporting your board will trust?
Require AI SEO, GEO, or digital marketing and AI automation within your business? Book a free appointment with a senior consultant.
FAQ
What should an SEO report include in 2026?
A modern SEO report should cover visibility (impressions for commercial queries), AI citations and share of voice from a fixed prompt panel, AI referral traffic from a dedicated GA4 channel, conversions, and pipeline from your CRM. Rankings and sessions still belong in the detail, but not as the headline.
How do I track ChatGPT traffic in GA4?
Create a custom channel group, add an “AI assistants” channel where Source matches a regex including chatgpt.com, and move it above Referral so it claims those visits first. ChatGPT adds utm_source=chatgpt.com to its links, which helps, but visits from the ChatGPT mobile app usually arrive as Direct.
What is AI share of voice?
AI share of voice is the share of AI answers that mention your brand, out of all answers that mention any brand you track, across a fixed set of prompts. It shows how visible you are in AI assistants compared with competitors. It doesn’t tell you whether the mentions are positive, so pair it with manual checks.
Why are my organic clicks down when my rankings haven’t changed?
Most likely because AI Overviews now answer more of the queries you rank for, so fewer searchers click any result. Ahrefs found a 58% lower CTR for the top-ranking page when an AI Overview appears. Check whether you’re cited in those Overviews, because cited brands earn noticeably more clicks than uncited ones.
Sources
- Ahrefs, “Update: AI Overviews Reduce Clicks by 58%” (February 2026)
- Pew Research Center, “Google users are less likely to click on links when an AI summary appears in the results” (July 2025)
- Seer Interactive, “AIO Impact on Google CTR: 2026 Update” (April 2026)
- Advanced Web Ranking, Google organic CTR study
- Ofcom, UK adults’ media and online lives (April 2026)
- Google Search Central, “AI features and your website”
- Google Analytics Help, custom channel groups
- OpenAI Help Center, Publishers and developers FAQ
- SearchPilot, “How does AI traffic show up in analytics?” (May 2026)
- Search Engine Land, Conductor AI referral benchmarks
- Ahrefs, “AI Share of Voice: What It Is, How to Measure It, and What It Actually Tells You” (October 2026)
- Ahrefs, Brand Radar
- Ahrefs, “Brand Radar Methodology: How we collect and model AI visibility data”


