Zero-Click in 2026: What the 68% Figure Actually Measures

Quick answer

The 68.01% figure means that, in Similarweb's mobile and desktop web panel observed in the United States between January and April 2026, slightly more than two Google searches out of three produced no measured click through to a result. It does not mean that 68% of users were satisfied, that 68% of queries display an AI Overview, or that sites lost 68% of their traffic. The study excludes the Google mobile app and models part of mobile behavior: it's a journey benchmark, not a universal rate.

Key takeaways

• Zero-click describes the absence of an observed click after a search, not the absence of value for the user or the brand.
• The 2026 rate of 68.01% comes from a US panel and explicit assumptions; it doesn't automatically apply to France.
• The 2024 and 2026 results use different vendors and weightings: the increase doesn't form a perfectly comparable series.
• A reformulated search is counted as click-free, but it reveals a different need than a session that simply ended.
• To decide, measure exposure, the available click, brand visibility, and final conversion separately.

Operating definition of zero-click

A search is called zero-click when the observed sequence contains no click on the results presented. The result can be organic, an ad, owned by Google, or point to the open web. Depending on the protocol, a new query or the end of the session closes the sequence.

That definition nonetheless aggregates several situations:

1. the user got an answer in the SERP;
2. they found no answer and gave up;
3. they reformulated the search;
4. they remembered a brand and navigated directly later;
5. a click or transition isn't visible to the measurement setup.

So the statistic distinguishes neither satisfaction, nor intent, nor delayed influence. It measures a technical event within a defined window.

Where does the 68.01% figure come from?

The SparkToro / Similarweb study published in June 2026 uses Similarweb's clickstream panel for Google searches performed in a browser in the United States, on desktop and mobile, from January to April 2026. The authors settled on a weighting of two thirds mobile searches and one third desktop. On mobile, ten seconds of inactivity marks the end of the session. Since mobile ad clicks aren't directly calculable in this dataset, they were estimated from an external ratio.

The study itself notes that the Google mobile app, which is likely to show many enriched answers, is absent. It doesn't publish the exact size of the sub-panel used in the article. The result is still useful because the calculation choices are described; it should be cited with those choices.

Why comparing it to 2024 calls for caution

In 2024, SparkToro and Datos studied a different panel, made up of tens of millions of users, on desktop and mobile. The data ran from September 2022 to May 2024, but the mobile component was only available from January to May 2024. The panel covered the United States and the European Union.

That study found 58.5% of searches with no click in the United States and 59.7% in the European Union. Around 37% of searches ended with the session and 22% with another search. For every 1,000 searches, 360 US clicks and 374 European clicks went to the open web. The panel covered iOS thinly, didn't control for ad blockers, and weighted mobile at 63% in the United States and 64.7% in the EU.

Comparing 58.5% in 2024 with 68.01% in 2026 gives +9.51 points and roughly 16.3% in relative terms. But that calculation mixes Datos and Similarweb, different panel compositions, different mobile rules, and different periods. It describes a directional signal, not an experiment measuring the isolated effect of AI Overviews.

Panel, queries, and visits: three denominators

Zero-click can be calculated from tracked people, from a keyword portfolio, or from visits to a SERP. The choice changes the result.

Pew tracked 900 American adults and 68,879 searches in March 2025. Around two thirds of all searches led either to somewhere else on Google or to the end of the session with no click on a result. That finding resembles the global phenomenon, but the definition and the period aren't Similarweb's. On pages with an AI summary, 8% of visits clicked a classic result and 1% clicked a source of the summary; with no AI summary, 15% clicked a classic result. The SERPs, however, were recreated from 7 to 17 April after the March browsing, and could have changed.

Procedure: estimating your economic exposure to zero-click

1. Segment queries by intent: information, comparison, navigation, transaction, and local.
2. Measure impressions and clicks over three comparable months in Search Console, by device and country.
3. Label exposure to surfaces with dated SERP observations; don't reapply today's label to the whole history without flagging it.
4. Calculate the CTR specific to each segment, not a global average.
5. Add the click-free signals: brand searches, direct visits, assisted conversions, and tracked citations, without adding them up as if they were the same unit.
6. Build three decline scenarios and define the threshold that genuinely changes the investment.

Reusable formulas

observed click-free rate = (searches with no click / observed searches) × 100

expected clicks = impressions × exposed share × exposed CTR + impressions × (1 - exposed share) × unexposed CTR

click loss = baseline clicks - expected clicks

Worked example: don't apply 68% to a site's impressions

Assume 200,000 monthly impressions, 60% of them informational. A survey shows that 50% of informational impressions trigger an enriched answer. The historical CTR is 3.5%; the scenario applies a 1.8% CTR on those queries. The other impressions keep 3.5%.

• Exposed impressions: 200,000 × 60% × 50% = 60,000
• Exposed clicks: 60,000 × 1.8% = 1,080
• Unexposed impressions: 140,000
• Unexposed clicks: 140,000 × 3.5% = 4,900
• Scenario total: 5,980 clicks, versus 7,000 at the historical uniform CTR

The modeled drop is 1,020 clicks, or 14.6%, not 68%. A panel's zero-click rate describes every search in the panel; it isn't a coefficient to subtract from a site's traffic.

The contradictory evidence on AI Overviews

The presence of an AIO and global zero-click are not the same thing. Semrush and Datos tracked more than 200,000 keywords with volume above 100 between January and October 2025. For the same terms with no AIO in May but with an AIO in October, the zero-click rate fell from 33.75% to 31.53%. That result doesn't prove the AIO creates clicks: other SERP changes and the intent mix aren't controlled. It shows the relationship isn't mechanical.

Conversely, Pew observes fewer clicks when the summary is present, and Ahrefs models a sharp drop in position 1 CTR. These results cover different units. The Observatory therefore holds a cautious conclusion: some surfaces answer without a click and shift how clicks are distributed, but the effect must be estimated by intent, interface, period, and portfolio.

What the data proves and doesn't prove

The studies prove that a large share of observed journeys after a Google search contains no click through to a result. They also document an increase between several snapshots, even if the exact measurement varies.

They don't directly measure satisfaction, brand recall, delayed visits, phone calls, store visits, or conversions that travel through another channel. They don't demonstrate that AI explains the whole change: featured snippets, maps, weather, local results, new queries, and Google properties also contribute to zero-click.

Common mistakes and stopping conditions

• Writing "68% of traffic has vanished": the denominator is the search, not a site's traffic.
• Presenting 58.5% then 68.01% on a continuous curve without flagging the panel change.
• Confusing the end of a session with satisfaction.
• Counting multiple clicks as unique searches.
• Using a US average to set a French budget.
• Stopping a forecast if the genuinely exposed queries aren't identified or if seasonality isn't controlled.

Reusable asset: the zero-click reconciliation sheet

For each figure, fill in this row: source | territory | dates | devices | app included? | denominator | session-end definition | clicks included | weighting | result | limitation. Two rates can only appear in the same trend if the columns are compatible; otherwise, draw a method break.

How SEOryon fits in

SEOryon can help segment Search Console data and keep dated observations where those functions are available. It cannot observe the searches where your site has no impressions, nor automatically attribute a direct visit to AI exposure. The platform should therefore serve the site's own scenario, while external panels provide the market context.

Measurable exercise

Take the 100 most exposed informational queries on your site. Deliver a table with country, device, impressions, CTR, dated AIO presence, intent, and a low, central, and high scenario. Success: the three scenarios recalculate with the formulas above and no external figure is applied without adjusting the genuinely exposed share.

FAQ

Does zero-click mean SEO is pointless now?

No. SEO is still needed to be discovered, cited, and chosen, but traffic is no longer the only output. Visibility, brand searches, visits, and revenue must be connected without being confused with each other.

Are AI Overviews responsible for the 68%?

The study doesn't establish that. AI Mode accounted for only 0.34% of observed searches in the January to April 2026 panel. Other SERP answers and behaviors already existed.

Which rate should you use for a French forecast?

Use your French Search Console CTR, segmented by intent and device, then apply a range to the observed share of exposed queries. The US 68.01% serves as context, not as a coefficient.

Is a reformulated search really click-free?

In these studies, yes, if no result was clicked before the new query. But analytically it deserves a separate category from a finished session, because it reflects an unresolved or refined need.

Read next

Return to the Observatory's methodological register, clarify the objectives with SEO, GEO, and AEO, then measure AI Overviews CTR. To move from an external benchmark to your own property, apply the guide to measuring AI visibility.

References

• SparkToro / Similarweb: 2026 zero-click study- SparkToro / Datos: 2024 United States and European Union study- Pew Research Center: clicks with AI summaries- Semrush / Datos: AI Overviews and zero-click- Google: AI features and websites

Method and update note

Reviewed 16 July 2026, translated and edited 22 July 2026. The 68.01% rate always travels with "Similarweb US web panel, January to April 2026." Any future edition must separately preserve values whose panel, device, or session rule changes.