68% Of Google Searches Now End Without A Click. AI Overviews Did Not Start That - They Finished It

68% Of Google Searches Now End Without A Click. AI Overviews Didn’t Start That — They Finished It

The number is doing the rounds and it deserves to. SparkToro’s analysis of Similarweb clickstream data found that **68.01% of US Google searches ended without a click** between January and April 2026.

Two thirds. Of every hundred searches, sixty-eight produce no visit to any website at all.

Most coverage of this figure has framed it as the AI Overviews story — Google built an answer engine, the answer engine ate the clicks, here we are. That framing is popular, it’s emotionally satisfying if you’ve spent a decade doing SEO, and it’s mostly wrong. What the data actually shows is more interesting and considerably less comfortable.

The number, and what it’s measuring

First, precision, because this stat gets mangled constantly.

The study covers **US Google searches, January to April 2026**, using Similarweb clickstream panel data across desktop and mobile. “Zero-click” means the search session ended without a click to any external result. It includes searches where the user found their answer in the SERP, searches that were abandoned, and searches that led to another search.

That last category is not a rounding error. Alongside the headline figure, the study found a **7.2 percentage point increase since 2024 in searches leading to another Google search** — people refining, not being satisfied. And a **9.51 percentage point decline** in searches producing at least one click.

So “68% zero-click” is not “68% of people got their answer from an AI summary.” It’s a broader behavioural measure. Anyone quoting it as pure AI cannibalisation is overstating.

The trend line before AI Overviews existed

Here’s the part that changed my mind about the whole narrative.

The historical series in the same research runs roughly: **~50% in 2019. 64.82% in 2020. 58.5% in 2024. 68.01% in early 2026.**

Look at 2020. Zero-click hit nearly 65% before a single AI Overview existed, before ChatGPT was public, before any of this. Featured snippets, knowledge panels, local packs, weather boxes, calculators, sports scores, shopping carousels — Google spent a decade building a results page that answers the query in place. The zero-click era was well underway.

AI Overviews didn’t create the trend. They accelerated a trend that had been running for years, and they did it by generalising something Google had previously only been able to do for structured queries. A featured snippet needed a page with the answer in a liftable format. An AI Overview can synthesise one from several sources for a query nobody has written a page about.

Why does the distinction matter? Because it changes the prognosis. If AI Overviews caused this, you could hope for regulation, backlash, or a product retreat. If AI Overviews are the latest expression of a fifteen-year strategic direction, the trend continues regardless of what happens to any particular feature.

I think it’s the second one, and I think planning on the first is how teams lose two years.

What the 60% CTR drop does to a category page

The other figure from the research: AI Overviews appear on **more than 20% of searches**, and where they appear, **click-through rates drop by nearly 60%**.

Make that concrete. A collection page ranking third for a decent commercial term, pulling 2,000 clicks a month. An AI Overview starts appearing on that query. Same ranking, same traffic to the SERP, roughly 800 clicks. You’ve lost 1,200 visits and nothing about your page changed.

Now the strategic question, and it’s the one I’d want a merchant to sit with. Those 1,200 people didn’t stop existing or stop wanting the thing. They got their answer somewhere in the summary. The question is no longer “how do I rank for this” — you already do. It’s “when that summary is generated, is my brand in it, and is my page the source.”

Those are different problems with different solutions. Ranking rewards authority, links and relevance. Citation rewards being the clearest, most specific, most extractable source of a particular fact — and being corroborated elsewhere.

AI Mode: small today, doubling quarterly

One more number, and it’s the one I’d watch.

AI Mode accounted for just **0.34% of searches** during the study window. Genuinely tiny. But it passed **1 billion monthly users by I/O 2026**, with query volume roughly **doubling each quarter**.

Take those two facts together carefully, because they’re in tension. Enormous user reach, minimal share of actual search volume — which tells you people are trying it, not living in it. That’s a product in the awkward middle: past novelty, short of habit.

Doubling quarterly is the part that matters. Four doublings from 0.34% is 5.4%. Six is over 20%. I’m not predicting either — compounding curves flatten and product bets fail — but the range of plausible outcomes eighteen months out spans “still marginal” to “a fifth of search,” and a strategy that only works in the first case is a bad strategy.

If not clicks, then what

The honest answer is that the industry hasn’t fully settled this, and anyone claiming otherwise is selling something. But three measures are holding up better than sessions:

**Citation share.** Across a fixed set of prompts and queries representing real customer questions, how often does your brand appear, and how often is your domain the cited source? Fixed prompt set, tracked monthly, benchmarked against named competitors.

**Quality per session, not sessions.** If total sessions decline while the surviving sessions convert better, that’s not necessarily a loss. Shopify’s Q1 2026 data found AI-referred sessions converting roughly **50% higher** than organic on product pages, with **14% higher AOV**. Fewer, better visits is a real outcome and your reporting should be able to represent it.

**Branded search volume.** The most underrated indicator in this environment. If people encounter you in a zero-click context and later search your name, that shows up. Branded search is becoming the closest available proxy for the impressions you can no longer count.

What to change in your reporting this quarter

Three things, none of them expensive:

**Stop reporting organic sessions as the headline.** Report organic revenue and organic conversion rate alongside it. A traffic decline with flat revenue is a story about zero-click, not a story about failure — but only if your report can show it.

**Add branded search volume as a tracked metric.** From Search Console, monthly, as a trend.

**Build a fixed prompt set and start tracking citation share now**, even crudely and manually. Thirty prompts phrased the way customers actually describe their problem, run monthly across the platforms that matter to you, recording whether your brand is mentioned and whether your domain is cited. You cannot retroactively create a baseline, and in twelve months the teams who started this quarter will be the only ones who can show a trend.

And one thing not to do: don’t rebuild your content strategy around chasing featured placement in AI Overviews specifically. The surface will change — it has changed twice already — and content written to game one presentation format ages badly. Write the clearest, most specific, most factually complete version of the answer, on the page that can be bought from. That has been the durable strategy through every SERP change of the last fifteen years, and there’s nothing in this data suggesting it stops working now.

The 68% is not a crisis number. It’s a measurement problem wearing a crisis costume. The demand didn’t vanish — the click did, and the click was only ever a proxy for the thing we actually cared about.

What’s genuinely uncomfortable is that we built an entire discipline, a set of tools and a way of reporting to management on top of that proxy. Replacing it is going to take a while, and the trend line isn’t waiting.

*Sources: SparkToro analysis of Similarweb clickstream data, US, January–April 2026, reported via Search Engine Land; Shopify Q1 2026 commerce data.*