Zero-click is 60%, or 22.4%, from one provider

TL;DR. Pew Research tracked 68,000 real search queries and found users clicked a result 8% of the time when an AI summary appeared and 15% when it did not, a relative reduction of about 47%, with 26% of summary-present sessions ending entirely against 16% otherwise. Reuters Institute and Chartbeat found Google search traffic to publishers down 33% globally in the year to November 2025, 38% in the US, and 60% over two years for publishers under 10,000 daily page views. Every method points the same way. And the most-quoted single statistic, the zero-click rate, is reported at 60% by one analysis and 22.4% by another using strict clickstream data, with the same data provider behind both. Direction is established. Magnitude is not, and this domain has no disclosure obligation at all.

---

Status: direction established, magnitude contested. Sources include Pew Research, the Reuters Institute with Chartbeat, and several commercial analytics providers. Disclosure: this site depends on search referrals and is subject to the phenomenon it describes, which is stated rather than left for a reader to infer.

---

What the strongest study found

The Pew Research Center tracked 68,000 real search queries from a panel, measuring what people actually did rather than modelling it from keyword tools.

Users clicked a result 8% of the time when an AI summary was present, and 15% of the time when it was not. That is a relative reduction of roughly 47%.

And 26% of sessions containing a summary ended the search entirely, against 16% for sessions without one.

Google disputed the methodology, arguing the analysis period overlapped with algorithm testing unrelated to AI summaries. That objection is worth carrying: it is specific, it is testable in principle, and it comes from the only party with the underlying data.

What the traffic data shows

The Reuters Institute and Chartbeat reported in January 2026 that Google search traffic to publishers fell 33% globally in the year to November 2025, with US publishers down 38% and Google Discover referrals to more than 2,500 publisher sites down 21%.

Chartbeat's March 2026 release found the damage concentrated by size: publishers with fewer than 10,000 daily page views saw 60% declines in search referral traffic over two years.

And the substitute did not arrive. ChatGPT referrals to publishers grew by more than 200% and still account for less than 1% of all referrals. Chatbot traffic is not offsetting search declines and is not close.

Other measurements agree on direction and differ on size. Ahrefs measured a 58% CTR reduction on the top-ranking page for keywords with AI summaries. Seer Interactive tracked organic CTR on informational queries falling from about 1.76% to 0.61%, a 61% drop. A randomised field experiment found summaries reduced organic clicks on triggered queries by 38%.

And the headline number falls apart

The statistic in general circulation is that around 60% of Google searches now end without a click, up from 58% in 2024, attributed to SparkToro with Datos.

Datos's own Q1 2026 State of Search report, using what it describes as a strict clickstream methodology, put the US zero-click rate at 22.4% in March 2026, down from 24.5% in December 2025.

Sixty percent and twenty-two point four percent, with the same data provider behind both.

Elsewhere the same metric is reported at 69% for the period May 2024 to May 2025, described as the sharpest year-on-year rise on record. And Semrush reported the zero-click rate specifically for summary-bearing keywords falling from above 45% in January 2025 to 38% by October.

These cannot all be measuring the same quantity. They differ on whether navigational and branded searches count, whether a click within Google's own properties counts as a click, whether the unit is a search or a session, and whether the panel is representative.

Nobody publishing these figures is obliged to state which choices they made, and most do not.

Which is Territory 8's finding again, in a domain with no regime at all

The previous territory closed on an observation: the reliability of a figure tracked whether anyone was required to publish it.

This domain is the extreme case. There is no securities filing, no regulatory return, no statutory obligation of any kind. Every number comes from an SEO tool vendor, an analytics provider, a publisher trade body, or a platform, and each of those has a commercial or institutional interest in the answer.

The exception is Pew, which measured behaviour with a panel, published its method, and had its result disputed by the platform. That is what the good case looks like here: one rigorous study, contested by the measured party, with no way for either side to settle it publicly.

And the platform holds the only complete data. Google knows exactly how many searches end without a click. It is not obliged to say and does not.

What is not in dispute

Worth separating, because the measurement mess can obscure it.

Traffic to publishers from search has fallen substantially, across every independent measurement, in every geography measured, over two years.

Smaller publishers are hit harder, which is the finding with the clearest mechanism: a large brand retains direct and app traffic, a small site was almost entirely search-dependent.

Chatbot referral has not substituted for it, at less than 1% of referrals despite very rapid growth.

And the effect concentrates on informational queries, the ones beginning "what is" and "how to", while transactional and navigational searches are less affected because the user still needs to arrive somewhere.

Three things this establishes

Direction and magnitude are separable claims and only one is settled. Every method agrees traffic is down. Estimates of the click reduction range from 38% to 61% depending on what is measured, and the zero-click rate varies threefold. Anyone quoting a single figure has selected one.

Concentration matters more than the average. A 33% global decline and a 60% decline for the smallest publishers describe the same phenomenon, and the second is the one that closes sites.

And the party with the complete data is the party being measured. That is the structural condition of this subject, it is not unusual, and it means every external figure is an estimate from a panel or a proxy no matter how carefully constructed.

What it does not establish

That AI summaries are the sole cause. Search algorithm changes, platform competition, shifts to social and video, and post-2021 traffic normalisation all overlap the same period, and no study isolates the summaries cleanly. The randomised field experiment is the closest thing to a clean identification and it covers one mechanism on triggered queries.

That the decline continues at this rate. Some measurements show early-2026 stabilisation, and Semrush reports the zero-click rate for summary-bearing keywords falling rather than rising.

That publishers are uniformly harmed. Brands cited within summaries reportedly earn more organic clicks than those not cited, so the effect redistributes as well as reduces.

And nothing about what should be done. Antitrust complaints and lawsuits are live in several jurisdictions and this corpus takes no position on them.

What is unresolved

What the zero-click rate actually is. It is the most quoted number in this subject and it is not established within a factor of three.

Whether the platform's methodological objection to Pew holds. It is specific and only the platform can test it.

Whether stabilisation is real or an artefact. Two independent measurements suggest the summary-specific zero-click rate is falling, which is either genuine adjustment or a change in what is being counted.

And what happens to the source material. If small publishers close, the corpus that summaries are generated from thins, and no measurement of that exists because it has not happened yet at scale.

The counter-argument

The zero-click comparison is unfair. Sixty percent and 22.4% are almost certainly measuring different populations, one including navigational and branded queries and the other excluding them, and presenting them as a contradiction from one provider is rhetorically effective and analytically loose. The provider is not contradicting itself; two studies with different scopes are being set side by side.

Search traffic was never an entitlement. Publishers built businesses on referral flows from a platform under no obligation to send them, and the decline is a change in a commercial arrangement rather than a harm in itself. That framing is contested and it is not obviously wrong.

The interest disclosure cuts against this article. A site dependent on search referrals writing about search referral decline has an interest in the decline being large and attributable. This one has tried to weigh the evidence and cannot claim to be disinterested.

And the small-publisher finding may be survivorship-shaped. Chartbeat measures sites that still exist and still run its analytics. Sites that closed are absent from the two-year comparison, which would make the measured decline an underestimate, or which could reflect composition change rather than uniform decline.

The short version

Pew tracked 68,000 real queries and found clicks at 8% with an AI summary against 15% without, a relative fall of about 47%, with 26% of summary sessions ending entirely against 16%. Google disputed the method, citing overlapping algorithm tests.

Reuters Institute and Chartbeat found publisher search traffic down 33% globally and 38% in the US in the year to November 2025, with 60% declines over two years for publishers under 10,000 daily page views. ChatGPT referrals grew over 200% and remain under 1% of all referrals.

And the headline statistic does not survive inspection. Zero-click is reported at 60%, at 69%, and at 22.4% on strict clickstream data, with one provider behind both the 60% and the 22.4%. They differ on whether navigational searches count, whether in-platform clicks count, and whether the unit is a search or a session, and nobody is obliged to say which they chose.

Which is the previous territory's finding in a domain with no disclosure regime whatsoever. Every figure comes from a tool vendor, an analytics firm, a trade body or the platform. The one rigorous panel study was disputed by the party it measured, and that party holds the only complete data and is not required to publish it.

What is not in dispute: traffic is down, smaller publishers are hit hardest, chatbots have not substituted, and informational queries take the damage. Direction is established. Magnitude is not, and this article has an interest in it, which is stated rather than concealed.

Common questions

How much has search traffic to publishers actually fallen? The Reuters Institute with Chartbeat found Google search traffic to publishers down 33% globally in the year to November 2025 and 38% in the United States, with Google Discover referrals to more than 2,500 sites down 21%. Chartbeat's March 2026 data found publishers with fewer than 10,000 daily page views down 60% over two years. The direction is consistent across every independent measurement; the magnitude varies by method.

What did the Pew study find? Tracking 68,000 real search queries from a panel, it found users clicked a result 8% of the time when an AI summary was present and 15% when it was not, a relative reduction of roughly 47%, and that 26% of summary-present sessions ended the search entirely against 16% otherwise. It measured behaviour rather than modelling it from keyword tools, which is why it carries more weight than most figures in this area. Google disputed the methodology, arguing the period overlapped with unrelated algorithm testing.

Why is the zero-click rate reported so differently? Because the measurements count different things. One analysis puts it at about 60% of Google searches, another at 69% for a 2024 to 2025 window, and a strict clickstream study puts the US rate at 22.4% in March 2026, down from 24.5% in December 2025. The same data provider sits behind both the 60% and the 22.4% figures. They differ on whether navigational and branded searches are included, whether clicks within the platform's own properties count as clicks, whether the unit is a search or a session, and how the panel is constructed. Almost none of these choices is stated alongside the figure.

Are AI summaries definitely the cause? Not established cleanly. Algorithm changes, competition from social and video platforms, and post-2021 traffic normalisation all overlap the same period. The closest thing to clean identification is a randomised field experiment finding summaries reduced organic clicks on triggered queries by 38%, which isolates one mechanism on one query type rather than explaining the aggregate decline.

Has chatbot traffic replaced search traffic? No, and not close. ChatGPT referrals to publishers grew by more than 200% and still account for less than 1% of all referrals. The substitution that would offset search declines has not occurred at any meaningful scale.

Why are smaller publishers hit hardest? Because they were more search-dependent to begin with. A large brand retains direct traffic, app users, newsletters and social distribution; a small site was often almost entirely reliant on organic search for discovery. Chartbeat's finding of 60% declines for sites under 10,000 daily page views describes that dependency being removed. The measurement may also understate the effect, since sites that closed are absent from a two-year comparison of sites that still exist.

Why does this subject have such poor measurement? Because nothing obliges anyone to publish. There is no securities filing, no regulatory return and no statutory reporting requirement, so every figure comes from an SEO tool vendor, an analytics provider, a publisher trade body or the platform itself, all of which have commercial or institutional interests in the answer. The platform holds complete data on how many searches end without a click, is not required to publish it, and does not.

Does this site have an interest in the answer? Yes, and it is stated for that reason. Artifipedia depends on search referrals and is subject to the phenomenon described here, which gives it an interest in the decline being large and attributable to AI summaries. The evidence has been weighed as carefully as possible and the article cannot claim to be disinterested, which is exactly the disclosure it asks of other sources.