The Google Traffic Panic: What the Data Actually Says

TL;DR

• Clicks on AI-answered queries have roughly halved. Position-1 click-through rate (CTR) falls 58% the moment an AI Overview appears on a search results page (Ahrefs, December 2025 data, 300,000 keywords).
• But AI referral traffic plateaued in mid-2025 and sits roughly ten to one behind Google. The chatbots are not where your traffic went.
• Over half of AI-influenced visits arrive as branded search, not a citation click. Your analytics logs that as "Search" or "Direct." It's neither. It's AI influence, invisible to your dashboard.
• Google now publishes AI impressions inside Search Console. Not clicks. You cannot verify Google's "no meaningful impact" claim. Neither can anyone else, including Google's critics.
• The right response is reweighting what you measure, not abandoning SEO. The strategy for doing that properly is the next piece in this series.

Two things about search are true right now, and most boardrooms are only hearing one of them.

When Google shows an AI Overview, people click through to a normal search result 8% of the time. When it doesn't show one, they click 15% of the time. That's Pew Research, built on 68,879 real Google searches from 900 US adults. Roughly half your click rate, gone, the moment an AI summary appears on the page.

Also true: AI chatbots send you roughly a tenth of the referral traffic Google does, per query, and that gap hasn't moved in well over a year.

Both numbers are real. Both come from credible, independently conducted research. If your AI search strategy is only built on one of them, you are about to spend budget in the wrong place, and probably already are.

That's the whole article in two sentences. Here's the data behind both halves, and what I'd actually do about it this quarter.

The numbers the panic is built on

Start here, because the panic isn't manufactured. The erosion is real, and the people telling you SEO is dying aren't lying to you. They're reading half the spreadsheet, not inventing one.

Pew Research tracked the actual browsing behaviour of 900 US adults through March 2025: 68,879 Google searches, 12,593 of which triggered an AI summary. The finding: users who encountered an AI summary clicked a traditional result in 8% of visits. Users who didn't, clicked in 15% of visits. Nearly double, in the other direction.

It gets worse before it gets better. Only 1% of users click a source link inside the summary itself. And sessions end entirely, no further click, no follow-up search, 26% of the time when a summary shows, against 16% when it doesn't. Google isn't just answering the question ahead of you. It's ending the session before you get a look in.

Ahrefs ran the sharpest version of this at scale: 300,000 keywords, comparing December 2023 against December 2025. Position-1 CTR on queries that trigger an AI Overview fell from 7.3% to 1.6%. Once you control for the general CTR decline across all search results over the same period, that isolates to a 58% reduction attributable specifically to the AI Overview itself, and Ahrefs called it their bleakest finding on the subject to date.

The drop isn't limited to position one, either. It softens further down the page, but it never disappears:

Seer Interactive, Kevin Indig, and Authoritas have all landed independently in the same 47 to 65% range, using different data and different methodologies. That's not one vendor with a story to sell. That's convergent evidence from three unrelated sources.

Zoom out further and it gets starker still. SparkToro's Rand Fishkin, working from Similarweb's clickstream panel, found 68.01% of US Google searches in early 2026 ended without a click to anywhere, up from 60.45% just two years earlier. For every 1,000 searches, only 276 clicks now reach the open web. In 2024, it was 374.

I went deeper on why CTR is dying as a metric for AI search, and what's actually replacing it, in Selection Rate Optimisation if you want the full mechanics.

The numbers the panic ignores

Here's the half of the spreadsheet almost nobody shows a Marketing Director.

Similarweb has tracked generative AI referral traffic across three consecutive reports. Between October 2024 and September 2025, monthly referrals from AI platforms climbed fast, from roughly 60 million to roughly 240 million. Real growth, and steep. But by late 2025 the line went flat, and it has stayed flat. Their most recent report confirms it directly: AI visits keep climbing toward 1.5 billion monthly, while AI referrals have sat stuck between 240 and 280 million since mid-2025. The two lines decoupled, and they haven't reconnected since.

Put a number on that gap. Google's referral rate held around 17 to 19% through mid-2025. AI Mode's referral rate sat between 1.6% and 2.5%. That's roughly ten to one. For every click Google sends outward, AI Mode sends a fraction of one. Chatbots are built to resolve the question inside the conversation, not to send you back out onto the open web.

So where does the commercial impact actually land? This is the finding that should be in every board deck right now, and currently isn't. Similarweb tracked real users for seven days after they received a ChatGPT brand recommendation, across Finance, Travel, and Beauty. When ChatGPT recommended American Express over Capital One, for instance, Amex saw more than double the visit rate of its unrecommended competitor within the week. Across the full study, brands recommended by ChatGPT were 2.5 times more likely to get a site visit within seven days. But here's the part that actually matters for your dashboard: 55.9% of that downstream traffic arrived via branded search, typed straight into Google, not through an AI referral link. Among users who weren't AI-influenced, branded search accounted for just 40.4% of visits. The AI channel itself, the bit that would actually show up tagged as a referral in your analytics, accounted for only 8.8% of the total effect.

Read that again. The majority of AI's commercial influence lands in your dashboard as Google branded search or Direct traffic. Your analytics were never missing this data. They were just crediting it to the wrong channel, every single time.

Now the myth that needs killing outright, because I hear it repeated in client meetings most weeks. Google announced at I/O 2026 that AI Mode had passed one billion monthly users, and a fair number of vendors have run with that to claim AI Mode is now the default search experience. It isn't. When a Chrome experiment briefly routed address-bar searches straight into AI Mode in June 2026, Google's VP of Search Engineering, Rajan Patel, called it an error on the record: "We're not planning to make AI Mode the default for Chrome searches." SparkToro's own data backs that up. Only 0.34% of searches transitioned into AI Mode across the same period that produced the 68% zero-click figure above. AI Mode is growing fast. It is not the thing eating your clicks. AI Overviews, present on roughly a fifth of all searches, are doing that.

The claim nobody can check

This is the sharpest point in the whole piece, so give it a moment before moving on.

When Pew published its findings, Google disputed them publicly, calling the methodology "flawed" and the queryset "skewed" and "not representative of Search traffic." Google's counter-claim is that AI Overviews send billions of clicks with no significant aggregate drop.

Then, on 3 June 2026, Google launched a new Generative AI performance report inside Search Console, giving site owners a dedicated view of AI Overview and AI Mode impressions for the first time. It's a genuinely useful addition, and I'd rather have it than not. It's also missing exactly one thing: clicks. No CTR. No queries. No split by surface. Impressions only, rolling out to a subset of UK properties first, alongside a new toggle that lets a site opt out of AI features entirely without touching its organic ranking. Useful control. Still no clicks.

Google is grading its own homework with data nobody else can see, while withholding the one metric that would actually settle the argument. I'm not accusing anyone of anything here. I'm pointing at the structure and letting you draw your own conclusion.

Practical takeaway: any vendor or agency telling you they can report your "AI CTR" right now is estimating it, modelling it, or making it up outright. Nobody has the raw number. Not them, not you, not Google's own critics. I'll get properly into why most AI visibility scores on the market are closer to noise than a usable metric in the third piece in this series.

What this means if you own a P&L

Stop measuring the old thing

Raw organic clicks stopped being a clean health metric the day AI Overviews rolled out across your top queries. Watch branded search volume trend instead, Google Search Console (GSC) impressions specifically on AI-eligible queries, and assisted or direct conversions on your key landing pages. I've written up the four-quadrant thinking behind which of your queries are even worth this level of attention in Grounding vs In-Model.

Do not fire your SEO. Rebrief them.

Google's own developer documentation is unambiguous on this: "You can apply the same foundational SEO best practices for AI features as you do for Google Search overall." Their newer optimisation guide, published May 2026, goes further still, telling site owners to explicitly ignore llms.txt files, content chunking, and AI-specific rewriting tactics. That's Google, in its own documentation, telling you the GEO hacks currently being sold to you as urgent are unnecessary. Worth remembering next time an agency pitches you on it.

The work doesn't stop. The scoreboard changes. What that scoreboard actually looks like, and how it serves Google rankings and AI visibility with a single integrated strategy, is the next article in this series.

Treat AI answers as a shop window you don't control

Being cited or recommended inside an AI answer influences buyers even when nobody clicks anything at all, and that influence lands in your analytics as branded search three days later, with the credit line stripped clean off it. Sixty seconds with the free Grounding Query Classifier will tell you which of your priority queries even route through an AI answer in the first place. Most site owners I run this with are surprised by how few actually do.

What I'd do this quarter

1. Pull twelve months of Google Search Console data. Segment out queries where impressions rose but clicks fell over the same window. That shape is the AI-answer signature, and it's usually easy to spot once you know what you're looking for.

2. Check branded search volume against that same window. Look for the displacement pattern sitting underneath it.

3. Classify your top 50 commercial queries as grounded or in-model before you spend a penny optimising a single one of them. The classifier does this in about a minute per query.

4. Set up GA4's native AI Assistant channel, live since 13 May 2026, and don't assume it catches everything on its own. Referrer-stripped visits from AI apps still land in Direct, so treat the number it gives you as a floor, not a ceiling.

5. Re-baseline what "good" looks like before the Q4 budget conversation. A falling click line, on its own, should not be allowed to drive a budget cut. It's telling you a fraction of the actual story, not the whole of it.

The honest bit

Every figure above is a July 2026 snapshot. Not a permanent map.

Pew's sample is US-only, and drawn from March 2025. Ahrefs' dataset skews toward their own customer base. Similarweb's figures come from panel modelling, not a full census of every search on earth. Google's aggregate claims are unverifiable in both directions, for the argument and against it. This picture will look at least somewhat different in six months. Anyone presenting AI-search data as settled science right now, me included, deserves a bit of healthy suspicion.

Where that leaves you

If your board is asking what the falling click line means and whether the SEO budget survives the next planning cycle, that's precisely the conversation my SEO advisory work exists for. If you need someone to own the answer day to day rather than just advise on it from the outside, that's fractional SEO director territory instead.

Next in this series: the single strategy that serves Google rankings and AI visibility at once, and why treating them as two separate disciplines is currently costing businesses real money.