AI Overviews Killed 60% of Clicks: Your 2026 Survival Plan

If your organic traffic charts started looking like a ski slope sometime in 2025, you already know what we're going to say. The AI Overviews click-through rate drop is real, it's measurable, and it's the single biggest reason ranking number one isn't paying out like it used to.

We run SEO campaigns for a living, so we've watched this happen across dozens of client accounts in real time. Same rankings, same impressions, fewer clicks. It's the most demoralising kind of decline because nothing looks broken. You're still winning the search result. Google just stopped sending the people.

So let's get into the actual numbers, who's getting hit hardest, where the panic is overblown, and what we'd genuinely do if it were our money on the line.

What is the AI Overviews click-through rate drop?

It's the gap between how often a search result used to get clicked and how often it gets clicked now that Google slots an AI-generated summary above it. The summary answers the question on the page, so the user has no reason to click through. Fewer clicks, same ranking.

The cleanest dataset on this comes from the Pew Research Center, which tracked 68,879 real Google searches from 900 US adults in March 2025. On pages with an AI summary, users clicked a traditional search result just 8% of the time. On pages without one, they clicked 15% of the time. That's close to half the clicks, gone, just because a summary appeared.

The link inside the summary? Barely touched. Pew found users clicked one of those cited links in only 1% of visits. So the consolation prize of "well, at least you might get cited" comes with a tiny asterisk we'll come back to.

The real numbers, from people who actually counted

There's a lot of made-up statistics floating around about this topic, so we'll only quote research we've read and verified ourselves.

The most rigorous keyword-level study is from Ahrefs, who analysed 300,000 keywords (150,000 with AI Overviews, 150,000 informational). Their finding, updated to December 2025 data: the presence of an AI Overview correlates with a 58% lower click-through rate for the top-ranking page. Their own framing is brutal and we love it for the honesty: for every 100 clicks a number-one page used to earn, Google now keeps 58.

What makes that figure land is the trajectory. Ahrefs' first study in April 2025 put the drop at 34.5%. Eight months later it was 58%. This isn't a one-off shock. It's a slope, and it's still tilting.

If you want the agency-side view, the data we trust most is from Seer Interactive, who looked at 3,119 queries across 42 organisations and 25 million organic impressions. Organic CTR on AI Overview queries fell from 1.76% to 0.61%, a 61% drop. Paid wasn't safe either, falling 68%, from 19.7% to 6.34%.

Here's the bit nobody quotes, and it's the most important one. Even on queries with no AI Overview at all, Seer found organic CTR fell 41%. The behaviour has changed everywhere. People are clicking less full stop, AI summary or not, because the whole results page now feels like a place to read rather than a place to leave.

So is your traffic actually gone? Not always

This is where we push back on the doom a bit, because the "AI killed search" headlines miss something real.

Search Engine Journal flagged a nuance from the same Seer dataset that more people should sit with. In one month, CTR fell hard while absolute clicks barely moved. Impressions doubled from 15.8 million to 33.1 million, and clicks went from 398,798 to 400,271. The CTR "collapse" was partly a maths illusion: impressions grew faster than clicks, so the ratio cratered even though the headcount held.

We see this in our own accounts constantly. A client panics because Search Console shows CTR halving, but their actual click count is flat or even up, because AI Overviews surfaced them for a pile of new queries they never used to appear for. Falling CTR with steady clicks is annoying, not fatal.

The month after, though, Seer's clicks did drop to 301,783 on even higher impressions. So our honest read: don't trust CTR alone, and don't trust a single month. Watch raw clicks and conversions over a rolling quarter. That's the number that pays the bills.

Which industries are getting hit hardest

The damage is not spread evenly, and it tracks neatly with one thing: how easily a query can be answered in two sentences.

The pattern we see is that purely informational, low-stakes queries get gutted. "What temperature to cook chicken", "how does compound interest work", "symptoms of a cold". Google can answer those completely, so the click never happens. If your content strategy is built on definitional and how-it-works articles, this is your bad year.

It's worse again in regulated spaces. Medical and health content sits right in the blast radius, which is why we've leaned hard into the higher standard Google holds YMYL and medical sites to with our healthcare clients. The bar to be the source that gets cited is genuinely higher there.

Who's relatively fine

Transactional and local intent survives because an AI summary can't complete the job. "Buy", "book", "near me", "compare prices", "demo". Google can describe a product, but it can't put it in your basket. Interestingly, Google's own AI Overviews on shopping queries have actually increased, and that's not all bad for retailers, because shopping intent still ends in a click.

B2B and considered purchases hold up too. Nobody signs a £40,000 software contract off the back of a three-bullet AI summary. They want case studies, pricing, a demo, proof. The deeper the consideration, the more the click survives.

Our blunt rule of thumb: if you sell the thing, you're mostly OK. If you only explain the thing, you've got work to do.

How to work out your own exposure in 4 steps

Before you change anything, measure. We do a version of this for every client whose traffic is wobbling, and you can run it yourself in an afternoon.

1. Pull your top queries from Search Console. Sort by impressions and clicks over the last 6 months. These are the keywords actually keeping the lights on.
2. Check which ones trigger an AI Overview. Search your top 30 to 50 terms in an incognito window and note which ones show a summary. That percentage is your raw exposure.
3. Compare CTR, not just clicks. In Search Console, filter to queries with AI Overviews and check whether their CTR has fallen while impressions rose. If clicks held steady, you're fine. If clicks fell on flat impressions, that's a genuine loss.
4. Tie it to revenue, not traffic. A 40% CTR drop on a query that never converts is noise. A 10% drop on your highest-intent commercial term is the one to fight for. Always weight by money.

If your traffic is sliding and you can't pin down why, this is exactly the kind of forensic work we do when a site is losing organic visibility. Half the battle is separating real loss from a scary-looking chart.

The citation question: worth it, or a vanity metric?

Everyone's chasing AI Overview citations now, so let's be straight about what they're worth.

On the upside, Seer found that sites cited inside an AI Overview got 35% higher organic CTR and 91% higher paid CTR than uncited sites in the same results. Being in the box clearly beats being below it. And there's a real branding effect: if your name keeps appearing in the summary, people start searching for you directly later, which is the one channel Google can't tax.

On the downside, remember Pew's 1%. The link inside the summary gets clicked almost never. So a citation is mostly an impression, not a visit. We treat citations as brand-building, not a traffic strategy, and we'd encourage you to do the same so you don't set yourself up for disappointment.

If you do want to compete for that slot, the mechanics are less mysterious than the noise suggests. We've written up exactly how to get cited in ChatGPT and AI Overviews, and the short version is: clear structure, genuine topical depth, and being the most quotable source on the page rather than the wordiest.

What actually still works in 2026

SEO didn't die. The cheap version of it did. The plays that still move the needle are the ones that were always the strongest, just with less room for error now.

Build real topical authority, not one good article

AI engines pull from sources they've decided are authoritative across a whole subject, not from a single lucky post. That means clusters: a pillar plus a network of supporting pieces that genuinely cover the topic. Strong internal linking patterns across your content are what tie that authority together and tell both Google and the LLMs that you own the subject.

Chase intent that requires a click

Shift weight towards commercial and transactional queries where the answer can't be summarised away. Comparison pages, product content, pricing, "best X for Y", local service pages. These still convert because the user has to leave Google to finish the task. The pure-information game is now mostly a citation game, so price it accordingly.

Treat AI search as its own channel

Being visible inside ChatGPT, Gemini and AI Overviews is a discipline of its own now, and it overlaps with classic SEO without being identical. This is genuinely the fastest-moving part of our work, and it's why we built out a dedicated AI search visibility service. If you want the strategic backdrop, we've laid out how to get your brand into AI answers in detail.

Stop treating Google as your only landlord

Relying entirely on one platform was always risky. With AI Overviews it's reckless. We've seen the survival playbooks up close, and the sites that took the smallest hits had diversified before the storm. The small publishers who lost 60% of their Google traffic and clawed it back all did the same things: built email lists, owned a community, leaned into video and platforms where AI hasn't eaten the answer yet.

Direct traffic, the kind where someone types your name in because they trust you, is the most valuable it has ever been. It's the one stream no algorithm update can confiscate.

Don't forget Google disagrees with all of this

For balance, and because it's genuinely funny, Google does not accept the premise.

When the Pew study landed, Google told Search Engine Land the research used "a flawed methodology and skewed queryset that is not representative of Search traffic", and insisted it still sends "billions of clicks to websites daily" with no significant drop in aggregate traffic. Their line is that AI features prompt people to ask more questions, creating new chances to reach websites.

Our take: both things are partly true, which is the uncomfortable bit. Aggregate clicks across the whole web can stay enormous while your specific informational pages quietly bleed out. "We send billions of clicks" is cold comfort when none of the missing ones were yours. Believe the data on your own queries over anyone's press statement, Google's included.

Where we'd put our energy if it were our site

Stripping out the noise, here's the honest priority order we'd work in.

• Measure first. Separate real click loss from CTR-with-steady-clicks. Don't fix a problem you don't have, and don't ignore one you do.
• Defend the money pages. Protect and double down on the commercial, high-intent queries that still demand a click. That's where revenue lives.
• Repurpose, don't abandon, your informational content. Turn it into the deep, quotable authority layer that wins citations and feeds your clusters, rather than expecting it to drive raw traffic on its own.
• Build channels Google can't tax. Email, brand search, community, video. Slowly, starting now.
• Get serious about AI visibility. Treat it as a real channel with real strategy, not a box to tick.

None of this is a magic recovery button. Anyone selling you one is having a laugh. But the sites we work with that started early aren't the ones panicking right now, and that head start is the whole game.

If your traffic charts have that ski-slope look and you'd rather not guess at the cause, that's literally what we do all day. Tell us what you're seeing and we'll help you work out whether it's a maths illusion or a genuine leak, and what's actually worth fixing.