A marketing manager opens Search Console and sees impressions up 41%, average position improved from 8.4 to 6.1, and clicks down 18%. Nothing has broken. Her agency has not stopped working. The numbers are telling her the truth. What they have stopped doing is meaning what they used to.
That pattern is now the normal state of a UK organic search report, and it is a reporting artefact as much as a performance signal. This piece explains why it happens, which of your existing metrics have quietly stopped working, and what to put in front of a board instead. I will call the replacement the Four Layer Board: surfaced, named, sought, earned.
Did our organic traffic just drop even though our rankings went up?
Probably, and the most likely cause is that Google is counting appearances inside AI features in the same bucket as ordinary blue links.
Since June 2025, AI Mode clicks, impressions and position have been folded into the Search Console Performance report inside the "Web" search type, with no filter to isolate them (Search Engine Land, 16 June 2025). Google's own documentation confirms the general rule: "Sites appearing in AI features (such as AI Overviews and AI Mode) are included in the overall search traffic in Search Console" (Google Search Central, AI Features and Your Website).
Follow the arithmetic. An appearance inside an AI Overview counts as an impression. Most of those impressions never produce a click, because the user got their answer. So your impressions rise, your clicks stay flat, and click through rate, which is one divided by the other, falls. Average position can improve at the same time, because AI feature placements sit high on the page. Four metrics move, three of them look like a problem, and your actual performance may not have changed at all.
This is the single most useful thing to explain to a board before you explain anything else. Otherwise every subsequent slide reads as an excuse.
When Google answers in an AI Overview, is that a win or a lost click?
Usually a lost click, sometimes a win, and the size of the loss is smaller than the headlines say.
The best independent evidence is from the Pew Research Center, which tracked real browsing behaviour for 900 US adults across 68,879 Google searches in March 2025. Users clicked a traditional search result on 8% of visits to a results page containing an AI summary, against 15% without one. Only 1% clicked a link inside the summary itself. And 26% of visits with an AI summary ended the browsing session, against 16% without (published 22 July 2025). Pew has nothing to sell, which makes it the cleanest source in this whole subject. It is also US only, from a 900 person panel, and mostly everyday informational search rather than commercial intent.
For the UK, the number to know is from SparkToro's analysis of Similarweb clickstream data for January to April 2026: 69.5% of UK Google searches ended without a click, the highest of the six countries measured, above the US at 68.0% and well above Germany at 62.1%. UK searchers also ended their session most often, at 43.1%, and sent 232 clicks to the open web per 1,000 searches against Germany's 287 (published 16 June 2026).
Seer Interactive adds the other missing half of the story. Across 53 brands and 5.47 million tracked queries, organic click through rate on AI Overview queries bottomed at 1.31% in December 2025 and recovered to 2.36% by February 2026 (published 24 April 2026). Being cited inside the Overview is worth 120% more organic clicks per impression than not being cited, though still around 38% below a query with no Overview at all. So the collapse narrative is roughly a year out of date, and being named in the answer is worth real money.
Which traditional SEO metrics are now misleading?
Three of them, and one has flipped from useful to actively harmful.
| Metric | Why it broke | What to use instead |
|---|---|---|
| Click through rate from Search Console | Denominator now includes AI feature impressions that were never going to produce a click | Clicks in absolute terms, plus impressions tracked separately as a visibility measure |
| Average position | Mixes blue link placements with AI feature placements in one average | Position for your money queries only, tracked manually or with a rank tracker that segments AI features |
| Total organic sessions | Now includes AI Overview and AI Mode traffic with no way to separate it, and loses AI assistant traffic that arrives with no referrer | Branded and non branded sessions split, plus the AI Assistant channel reported separately |
| Keyword rankings as the headline number | Still real, but decoupled from traffic | Rankings as a diagnostic, never as the top line |
Click through rate is the harmful one. A team that manages to click through rate will start avoiding the queries where AI Overviews appear, which are frequently the informational queries that build the brand familiarity that gets you named in AI answers in the first place. You would be optimising away from the thing that works because the ratio looks bad.
How do you measure visibility inside AI search features?
Partly with Google Search Console, mostly with sampling tools, and it is worth being blunt that none of it tells you what the visibility earned.
There are plenty of other KPIs you can track, and I would leave most of them out of a board pack, because a long list of AI metrics is how a report stops being read.
The six that matter
| KPI | What it tells you | Where you get it | Frequency |
|---|---|---|---|
| Share of voice | The percentage of your frozen prompt set where the engines name your brand. The headline number, and the one to trend | Your own prompt test, or an AI visibility tool | Monthly |
| Share of voice by engine | The same figure split across ChatGPT, Perplexity, AI Overviews and Gemini, which diverge enough that one blended number hides the story | Same | Monthly |
| Citation rate | How often your domain appears as a listed source rather than an unattributed mention. This is the half that can send traffic | Your own prompt test, or a visibility tool | Monthly |
| AI feature impressions | How often you surfaced inside an AI Overview or AI Mode. The only first party number in the set | Search Console, Search Generative AI report | Weekly |
| Linked versus unlinked mentions | How many mentions are clickable. Unlinked ones still count commercially and will never appear in your analytics | Reading the answers yourself | Monthly |
| AI crawler hits | Whether GPTBot, PerplexityBot, ClaudeBot and Google-Extended are actually reaching your pages | Server logs or your CDN | Monthly |
What is "share of voice" and how do you calculate it?
Share of voice in AI search is the percentage of a defined set of prompts where your brand gets named in the answer. You calculate it by fixing the prompt set, running it, and counting.
The method, in five steps:
- Write 40 to 60 prompts a real buyer would type, in the words they would use. Include the "best X in your city" shapes, the comparison shapes and the problem shapes.
- Freeze that list. If you change the prompts, the trend is worthless.
- Run every prompt across the engines that matter to you, at the same time of day, from the same location.
- Record whether your brand is named, whether it is linked, and which competitors appear alongside you.
- Divide your mentions by the total prompts run. That is your share of voice.
Run it monthly. The absolute number is close to meaningless, since a different prompt set gives a different answer. The direction of travel over six months is the whole point, and it is the one AI metric a board can read without a translation.
Frequently asked questions
Do AI Overviews really reduce clicks?
Yes, though less dramatically than the headlines suggest. Pew found users clicked a result on 8% of visits with an AI summary against 15% without (July 2025). But Ahrefs’ own control group of keywords with no Overview fell almost as far over the same period, so Overviews accelerated a decline that was already happening.
Should I still care about keyword rankings in 2026?
Yes, as a diagnostic, no longer as your headline number. Rankings still tell you whether your content is competitive on a query. They no longer predict traffic reliably, because a page can rank first and lose the click to an AI Overview above it. Report rankings underneath outcomes, never above them.
How do I prove SEO ROI to my board without click growth?
Show four layers rather than one: where you were surfaced, where you were named, whether branded search grew, and what it earned. Branded query volume and self reported enquiry source carry most of the weight. Pair them with a frozen prompt set measured monthly so the trend is comparable.
What tools track AI Overview and LLM visibility?
Ahrefs Brand Radar, Semrush’s AI Visibility Toolkit, Profound, Peec AI and Otterly are the ones named most often. All of them sample prompts rather than reading real query logs, so their absolute numbers differ from each other. Pick one, keep it, and report the trend.
Want this done on your site?
Tell me what you sell and where. I reply within 48 hours with a time to talk and a first read on what would move the needle, even if you end up doing it yourself.