There is a conversation happening in marketing meetings all over the country right now, and it goes something like this.
Organic traffic is down. Not a wobble, a proper decline. Twenty percent, thirty, sometimes closer to half. Someone pulls up the analytics dashboard and the line goes one way only. The obvious conclusion follows quickly: something has broken. A penalty, a botched migration, a Google update that quietly buried you.
Then someone checks the revenue numbers, and the story stops making sense. Leads are flat. Conversions are flat. Sales are flat. Nearly half the traffic has gone and the commercial engine has barely coughed.
If that sounds familiar, your SEO is probably not broken. Buyer behaviour is. And the numbers you have used to measure marketing performance for the last fifteen years are quietly measuring the wrong thing.
The metric that tells you the truth
Before you commission an audit, open Google Search Console and compare the last twelve months against the twelve before it. You are looking at two lines, not one.
Clicks are what your analytics platform reports as organic sessions. Impressions are how often Google showed your content to someone.
When traffic falls because your SEO has genuinely failed, both lines fall together. You lost rankings, so Google stopped showing you, so nobody clicked. That is a visibility problem and it needs fixing.
When traffic falls because search behaviour has changed, the lines separate. Impressions hold steady or even climb, while clicks drop away underneath them. Google is still showing your content to as many people as ever. They are simply reading the answer without visiting you.
That divergence has a name in the industry now, the great decoupling, and it is no longer an edge case. It is the default.
The numbers behind the pattern
This is not agency folklore. The independent data is unusually consistent.
Pew Research Center tracked the real browsing behaviour of 900 US adults across nearly 69,000 Google searches. When an AI summary appeared in the results, users clicked through to a website on 8% of visits. When no summary appeared, they clicked on 15%. Roughly half the click-through rate, on the same searches, purely because the answer was already on the page. Only 1% clicked a link inside the AI summary itself. Searches ending the browsing session entirely rose from 16% to 26%.
Ahrefs studied 300,000 keywords, comparing performance before and after AI Overviews rolled out. Their finding was a 34.5% reduction in clickthrough rate for the top-ranking page when an AI Overview is present. Same position, same content, a third fewer clicks.
Seer Interactive's ongoing study of 5.47 million queries and 2.43 billion impressions across 53 brands shows the decoupling at scale, and shows that it moves. Through 2025, click-through rate on queries with an AI Overview compressed steadily to a floor of 1.3% in December, and the mechanism is the important part: impressions roughly doubled while clicks stayed broadly flat. Early 2026 brought a genuine rebound, with AIO click-through rate climbing back to 2.4% by February.
That last point matters, and it is where a lot of commentary published last year is now out of date. The floor was not permanent. But even after the recovery, queries with an AI Overview convert to clicks at 2.4% against 3.8% for queries without one. The gap has narrowed, not closed.
Read those three studies together and the conclusion is hard to avoid. A significant portion of the traffic that has left your website did not go to a competitor. It stopped being a website visit at all.
Why the revenue holds when the traffic does not
Here is the part that surprises people, and it is the most commercially important thing in this article.
The traffic you lost and the traffic you monetise were never the same traffic.
Think about what an AI Overview or a ChatGPT answer is genuinely good at. It handles the top of your funnel with total confidence. What is a fractional CFO. How does EPC certification work. What is the difference between a CRM and a marketing automation platform. These are definitional questions with settled answers, and a language model can resolve them in two sentences without anyone visiting a website.
Those queries used to drive enormous volumes of traffic. They also converted at close to nothing. They inflated your session count, propped up your dashboards and made your traffic graph look wonderful, while quietly contributing almost nothing to pipeline.
Now consider what a buyer does when they are actually choosing a supplier. They want to see your pricing. Your case studies. The people who would run the account. Whether you have handled a business that looks like theirs. Whether you feel credible. No AI summary settles that, because the decision is not informational, it is a judgement about risk and trust, and it requires the buyer to come and look at you.
So the informational traffic evaporates and the commercial traffic stays. Which is exactly why the graph collapses and the revenue does not.
Watch for the supporting signals in your analytics. When this pattern is playing out properly, average session duration usually rises and bounce rate falls. Fewer people arrive, but a far higher proportion of them arrive with intent.
What about AI referrals?
The tidy explanation would be that all those missing clicks simply moved to AI platforms, showing up as referrals from ChatGPT, Gemini or Perplexity instead.
The data does not yet support that, and you should be sceptical of anyone selling you a strategy built on the assumption that it does.
AI referral traffic is real and growing fast. Similarweb records around 9.5 billion monthly visits across generative AI platforms, up 70% year on year, and the rate at which ChatGPT cites external sources climbed from 1.6% to 6.8% of responses in twelve months. But for most businesses, AI referrals still represent a low single-digit percentage of total sessions. Nowhere near enough to account for a 40% drop in organic traffic.
The clicks did not move. Most of them ceased to exist.
What the AI traffic lacks in volume it makes up for in quality. Semrush's research puts the average visit from an LLM at 4.4 times the value of the average visit from traditional organic search when measured by conversion rate, and projects AI channels driving economic value comparable to traditional search by late 2027. A small stream of unusually well-qualified people, growing quickly. Worth building for, but not yet worth reporting as your primary channel.
Your five-minute diagnostic
Open Search Console, set the comparison to the last twelve months against the previous period, and match your data to one of these three patterns.
Impressions flat or up, clicks down, conversions steady. Your visibility is healthy and your content is still being surfaced. What changed is how people consume it. This is not an SEO failure and it does not need an SEO rescue. It needs a change in how you measure and where you invest.
Impressions down, clicks down, conversions down. You have a genuine visibility problem. Rankings, technical health, indexation, content quality or a competitor who has simply outworked you. This one warrants a full audit and it warrants it soon.
Impressions up, clicks down, conversions also down. The most nuanced case. Your content is being surfaced, but for the wrong queries, or the pages that used to convert are the ones losing clicks. Look at which specific URLs and query types are declining. If your commercial and bottom-funnel pages are the ones bleeding, that is a different and more urgent problem than losing your glossary traffic.
Most businesses we speak to fall into the first category and have spent months worrying that they are in the second.
What to do once you know which pattern you are in
If you are in the first pattern, three things change.
Change what you report. Sessions as a headline KPI has quietly become misleading, because it now measures a behaviour that is disappearing rather than a business outcome that matters. Lead with impressions and share of voice for visibility, conversions and pipeline for performance, and treat clicks as what they now are, one route among several by which your expertise reaches a buyer. If your board reporting still opens with a traffic graph, it is telling them a story about the internet rather than a story about your business.
Change what you publish. The commodity explainer, the definitional listicle, the thin how-to guide, these have lost their commercial function. Not because Google has punished them, but because an AI can now write and deliver that answer better and faster than your buyer can find your version of it. What cannot be replicated is original data, genuine practitioner experience, opinions with a name attached, and the specific detail of work you have actually done. That is also, not coincidentally, exactly the material that language models cite when they answer questions in your category.
Change what you optimise for. Being the source an AI system quotes is now a meaningful acquisition channel in its own right, and it is earned differently to a blue link. Clear structure, direct answers to real questions, strong entity signals, consistent information about who you are and what you do, and enough authority that a model treats you as worth citing. If you want the detail on that, our piece on why LLM optimisation is your next SEO superpower covers the mechanics, and AEO says your SEO playbook is obsolete covers the strategic shift.
The uncomfortable bit
A lot of marketing teams are about to spend serious budget fixing a problem they do not have.
The declining traffic graph is real, and it is frightening if sessions are the number you are judged on. But the graph on its own cannot tell you whether you have lost visibility or simply lost the clicks that were never worth much. Only the relationship between impressions, clicks and conversions can do that.
Check your impressions this week. If they are holding while your clicks fall away, you are not losing. You are being measured by an instrument that has stopped describing reality.
The businesses that work this out early get to stop panicking and start building for how buyers actually behave now. The ones that do not will spend 2027 optimising for a click that is not coming back.
Not sure which pattern your data is showing? We run a free visibility and search performance review that reads your Search Console and analytics together, and tells you plainly whether you have an SEO problem or a measurement problem. Book a consultation with Pathway.
Sep 29, 2026, 5:27:04 PM
