AI Search & Referral Traffic 2026

AI Search Traffic Loss Statistics 2026

How much referral traffic publishers have lost to AI search, how the losses break down by site size, what share of searches now end without a click, and how little traffic AI platforms send back. Every figure carries a named source and a date, and our own calculations state their inputs.

📊 45+ sourced statistics 📈 Similarweb, Chartbeat & Pew data 🔍 Breakdown by publisher size 🔄 Updated August 2026
Our analysis

Small publishers have lost nearly three times as much as large ones

The story is usually told as a single number for “publishers.” It is not one story. Chartbeat’s data, broken out by site size, shows the decline in search referral traffic over two years scaling almost inversely with how big the site is.

−60%search referrals to small publishers, over two years
−47%medium publishers, 10,000–100,000 daily pageviews
−22%large publishers, over 100,000 daily pageviews
See the full publisher breakdown →

Original analysis by Paid Hosting · August 2026 · small sites hit 2.7× harder than large · methodology · full source list

Key Statistics at a Glance

The headline figures on AI search and publisher traffic in 2026. Every number links to a named source in the sources list.

68%of Google searches now end without a click to any websiteSimilarweb · Jun 2026
8%of searches with an AI summary produced a click, against 15% without onePew Research Center · Jul 2025
<1%of all publisher pageview referrals come from AI chatbotsChartbeat via Axios · 17 Mar 2026
20%of people globally trust answers from AI chatbots, against 37% who trust news overallReuters Institute · Jun 2026
15 minof every hour spent looking for information online is now spent on the open webCloudflare · Jul 2026
−34%fall in publisher pageviews from Google Search between December 2024 and December 2025, with Google Discover down 15%Chartbeat via Axios · 17 Mar 2026

Zero-Click Searches

A zero-click search is one that ends without the user clicking through to any website. The share has risen steadily for a decade and sharply since AI Overviews arrived.

Share of Google searches ending without a click

Measured by clickstream panels. Panel provider changes over the period — see the caution below.

2016 ~45%
2019 ~49%
2020 ~65%
2024 ~60%
2026 68.01%

Source: Similarweb, 10 June 2026, compiling Jumpshot (2016, 2019), Similarweb (2020, 2026) and Datos (2024) clickstream panels.

Chart may be reproduced with attribution to paidhosting.com
  • 68.01% of all Google searches now end without a click to any website.Source: Similarweb clickstream panel, 10 June 2026, shared with Rand Fishkin of SparkToro for his June 2026 zero-click study
  • Around 45% of Google searches ended without a click a decade ago, making the increase roughly 23 percentage points over ten years.Source: Similarweb, 10 June 2026
  • AI Overviews now appear on more than 20% of all Google queries, and that share is still growing.Source: Similarweb, 10 June 2026
  • The share of searchers who perform another search rather than clicking to an external site rose 7.2 percentage points between 2024 and 2026.Source: Similarweb clickstream data, 10 June 2026
  • AI Mode still accounted for only 0.34% of all Google searches between January and April 2026 — meaning the next phase of zero-click growth has barely started.Source: Similarweb, 10 June 2026
⚠️
The 2024 dip is a panel change, not a real decline

The timeline above shows zero-click rates falling from about 65% in 2020 to about 60% in 2024, then rising to 68% in 2026. Searches did not become more clickable in 2024. That year’s figure comes from a different clickstream panel (Datos) than the surrounding years, and Similarweb flags it explicitly as less accurate. The reliable signal is the long arc from roughly 45% in 2016 to 68% in 2026. If you are quoting a single year-on-year change from this series, check which panel produced each end of the comparison — several widely shared charts do not.

The AI Overviews Effect on Clicks

Pew Research Center’s study remains the most methodologically transparent measurement of what an AI summary does to click behaviour, because it observed actual browsing rather than asking people to recall it.

  • Users who encountered an AI summary clicked a traditional search result in 8% of visits, against 15% for users who did not see one — a relative fall of about 47%.Source: Pew Research Center, 22 July 2025 (browsing data from 900 U.S. adults, March 2025); relative change calculated by Paid Hosting
  • Users clicked a link inside the AI summary itself in just 1% of visits to pages containing one.Source: Pew Research Center, 22 July 2025
  • 26% of visits to a search page with an AI summary ended the browsing session entirely, against 16% without — a relative increase of about 63%.Source: Pew Research Center, 22 July 2025; relative change calculated by Paid Hosting
  • 58% of respondents encountered at least one AI-generated summary in their Google searches during March 2025.Source: Pew Research Center, 22 July 2025
  • Wikipedia, YouTube and Reddit collectively accounted for 15% of the sources listed in AI summaries and a similar 17% of sources in standard search results.Source: Pew Research Center, 22 July 2025
📌
Falling clicks and rising ad revenue are not a contradiction

Referral traffic to publishers is falling sharply while total search volume is not. The mechanism connecting them is what Similarweb calls the repeat search loop: its clickstream data shows the share of searchers who perform another search rather than clicking out rose 7.2 percentage points between 2024 and 2026. Faster answers make users search more often, and those searches resolve on the results page. The traffic is not disappearing from the internet — it is being retained inside the search results page. Any analysis that treats declining publisher traffic as evidence of a shrinking search business has the relationship backwards.

Publisher Traffic Losses by Site Size

Chartbeat’s data, covering thousands of publisher sites globally, is the clearest available breakdown of who has actually lost traffic. The pattern is strongly size-dependent.

Fall in Google Search referral traffic by publisher size

Percentage decline in referral traffic from traditional search engines over two years, to March 2026. Size measured by average daily pageviews.

Small (1,000–10,000 daily) −60%
Medium (10,000–100,000) −47%
Large (over 100,000) −22%

Source: Chartbeat data provided exclusively to Axios, 17 March 2026. Chart compiled by Paid Hosting.

Chart may be reproduced with attribution to paidhosting.com
  • Referral traffic from traditional search engines declined 60% for small publishers over the two years to March 2026, against 47% for medium-sized publishers and 22% for large ones.Source: Chartbeat data provided exclusively to Axios, 17 March 2026
  • Small publishers were therefore hit roughly 2.7 times harder than large publishers.Source: Paid Hosting calculation from Chartbeat figures (60 ÷ 22)
  • Small publishers are defined as sites averaging 1,000–10,000 daily pageviews, medium as 10,000–100,000, and large as more than 100,000.Source: Chartbeat via Axios, 17 March 2026
  • Publisher pageviews from Google Search fell 34% and from Google Discover 15% between December 2024 and December 2025 — the two largest referrers of traditional search traffic globally.Source: Chartbeat via Axios, 17 March 2026
  • Average weekly pageviews across all global publishers measured by Chartbeat dropped 6% between 2024 and 2025. Chartbeat attributes this to normal variation outside search trends, including an off year for elections and a shifting news cycle, rather than to AI.Source: Chartbeat via Axios, 17 March 2026
  • Search losses at news and media sites are being partially offset by direct traffic and internal traffic from other pages on the same site, with referrals from email, apps and instant messages also growing as a share of overall traffic.Source: Chartbeat via Axios, 17 March 2026
  • Chartbeat aggregates this data from thousands of client websites globally and has tracked traffic trends across its network for nearly two decades.Source: Chartbeat via Axios, 17 March 2026
  • Some heavily crawled site categories have seen human traffic fall by as much as 40% in under a year.Source: Cloudflare, 1 July 2026
  • Google still accounts for approximately 88% of referral traffic, making it simultaneously the largest source of visitors and the largest substitute for them.Source: Cloudflare, 1 July 2026
💡
Why smaller sites lose more

The size gradient is the most useful finding in this data and the least reported. Axios’s own reading is that larger publishers with better brand recognition and stronger direct-to-consumer products are more insulated from declines in traditional search. A second mechanism is plausible but not established by the data: smaller sites skew towards informational and definitional queries, exactly what AI Overviews answer directly, while larger publishers hold more branded, navigational and breaking-news traffic that a summary does not replace. The practical implication either way is that aggregate publisher figures understate the damage to independent sites by a wide margin, and any conclusion drawn from a single blended number will be wrong for most of the web.

How Much Traffic AI Platforms Send Back

The obvious question is whether AI platforms are replacing the traffic that search no longer sends. On the published evidence, not at anything close to the scale of the loss.

  • AI chatbots account for less than 1% of all publisher pageview referrals.Source: Chartbeat data, reported by Axios, March 2026
  • ChatGPT referrals grew by more than 200% over the course of 2025 — substantial growth from a very small base.Source: Chartbeat data, reported by Axios, March 2026
  • Google AI Mode generates a referral on just 1.6–2.5% of queries, against 17–19% for traditional Google search — roughly seven to twelve times fewer.Source: Similarweb, 10 June 2026; ratio calculated by Paid Hosting
  • AI platform visits grew 76% year on year worldwide by the second half of 2025, yet referrals from those platforms plateaued over the same period.Source: Similarweb, 10 June 2026
  • News and media sites receive the highest overall number of pageviews from AI platforms but show the lowest engagement, consistent with users clicking through mainly to fact-check.Source: Chartbeat data, reported by Axios, March 2026
  • For every hour spent searching for information online, only about 15 minutes is now spent on the open web.Source: Cloudflare, 1 July 2026
  • More than 50 publisher–AI licensing agreements have been signed since 2023 as publishers seek to replace lost referral revenue.Source: Cloudflare, 1 July 2026

What Audiences Say They Do

The figures above measure what happens on networks and in browsers. The Reuters Institute’s Digital News Report surveys people directly, across 48 markets and roughly 97,520 respondents. It is the largest non-commercial dataset on news consumption, and it both corroborates the traffic decline and complicates the picture.

  • Google organic search traffic to more than 2,500 news sites fell 33% globally between November 2024 and November 2025, and 38% in the United States.Source: Chartbeat data, published in Reuters Institute Digital News Report 2026, 16 June 2026
  • Publishers expect traffic from search engines to almost halve, falling a further 43% over the next three years.Source: Reuters Institute Digital News Report 2026, 16 June 2026
  • Weekly use of AI chatbots for news rose from 7% to 10% globally, and reaches 16% among under-35s — fast growth, but not the explosion often described.Source: Reuters Institute, 16 June 2026
  • Trust in answers from AI chatbots stands at 20% globally, against 37% for news overall and 22% for news on social media.Source: Reuters Institute, 16 June 2026
  • Search as the main gateway to news is unchanged among older audiences and down only 3 percentage points among under-35s — a much smaller shift than referral traffic data implies.Source: Reuters Institute, 16 June 2026
  • Social media and video networks became the most widely used route to online news for the first time, at 54%, ahead of news organisations’ own websites and apps at 51% and TV at 52%.Source: Reuters Institute, 16 June 2026
  • Use of news websites and apps has fallen 12 percentage points since 2020, almost matching television’s 13-point decline over the same period.Source: Reuters Institute, 16 June 2026
  • Hard news search queries mostly remain excluded from AI Overviews, which is why news-specific impacts differ from the all-query figures reported elsewhere on this page.Source: Reuters Institute, 16 June 2026
  • Google Discover traffic to Reach, a listed UK news publisher, fell 46% in the second half of the year, according to its chief executive.Source: Piers North, quoted in Press Gazette, March 2026, cited in Reuters Institute, 16 June 2026
📌
Two independent readings of the search decline agree

Chartbeat’s data reached the public twice, through different outlets, covering slightly different windows. Axios reported publisher pageviews from Google Search down 34% from December 2024 to December 2025. The Reuters Institute published a 33% global decline across more than 2,500 sites from November 2024 to November 2025, with the United States at 38%. Two independent publications, one commercial newswire and one academic institute, landing within a percentage point of each other on overlapping windows. That convergence is the strongest evidence on this page that the search referral decline is real and correctly measured, rather than an artefact of one vendor’s panel.

⚠️
People say they click through far more than they actually do

The Reuters Institute found that 42% of people who use AI chatbots for news say they always or often click through to the original source. Pew Research Center, observing real browsing rather than asking, found users clicked a link inside a Google AI summary in 1% of visits, and clicked any traditional result in 8% of visits where a summary appeared. These measure related but not identical things — standalone chatbots versus AI summaries in search — and the Reuters Institute itself cautions that this is an area where stated and actual behaviour may diverge. The gap is nonetheless far too large to be explained by definitions alone. Survey data on click-through should be treated as a measure of intent, not behaviour. Any figure describing how often people click through to sources should say whether it was observed or self-reported, and most reporting does not.

Referral Quality: Fewer Visits, Better Visitors

The volume story is unambiguous. The quality story runs the other way, and it is the part most often left out.

  • Visitors referred by ChatGPT spent an average of 15 minutes on site, against 8 minutes for Google-referred visitors — roughly 1.9 times longer.Source: Similarweb 2025 Gen AI Landscape report, US desktop, September 2025; ratio calculated by Paid Hosting
  • ChatGPT-referred visitors viewed 12 pages per session against Google’s 9.Source: Similarweb, September 2025
  • On transactional sites, ChatGPT referrals converted at 7% against Google’s 5%.Source: Similarweb, September 2025
  • At the discovery and initial-ideas stage, 35% of US consumers rated AI tools as most useful, against 13.6% for search engines.Source: Similarweb 2026 AI Brand Visibility Report, January 2026 US consumer panel
  • A third of all ChatGPT citations in September 2025 pointed to URLs at folder depth three or deeper — specific interior pages rather than homepages or category hubs.Source: Similarweb, September 2025

Why Desktop and Mobile Tell Different Stories

The widely quoted 68% figure is a weighted blend of two very different surfaces, and the blend hides a much higher number.

  • On mobile, 45.2% of searches result in an external click and only 11.5% end in pure session abandonment.Source: Similarweb, 10 June 2026
  • On desktop, only 20.4% of searches produce an external click to a non-Google site, and 48% end with no activity at all.Source: Similarweb, 10 June 2026
  • Desktop zero-click exposure is therefore closer to 80% than to the headline 68%.Source: Paid Hosting calculation from Similarweb figures (100 − 20.4)
  • The 68% headline weights mobile at two-thirds and desktop at one-third, reflecting overall Google usage. Mobile is the more click-friendly surface, which is what pulls the blended figure down.Source: Similarweb, 10 June 2026
  • Informational, definitional and “how does” queries now resolve at zero-click rates well above 70% in most categories, while transactional queries with explicit purchase intent are far less affected.Source: Similarweb, 10 June 2026
  • Individual factual queries show the ceiling: “grams to pounds” carries a 100% zero-click rate, “weather” 85% and “capital of Sweden” 77%.Source: Similarweb keyword database, 10 June 2026
📌
Which 68% you are quoting matters

If a site’s audience researches on desktop — B2B software, professional services, technical documentation — the applicable zero-click rate is closer to 80%, not 68%. If the audience is mobile-first consumer, the rate is nearer 55%. Quoting the blended figure at either audience misstates their exposure by more than ten percentage points in opposite directions. The blend is correct for describing Google overall and misleading for describing any specific publisher.

Crawled Far More Than Referred

The other half of the exchange: AI systems are reading publisher content at enormous volume while sending back almost nothing. Cloudflare’s crawl-to-refer ratios quantify the imbalance directly.

  • Anthropic’s bots requested roughly 1,200 pages for every visitor referred back; Perplexity 746 pages, OpenAI 256.8 and Google 5.1.Source: Cloudflare Radar, worldwide, 2–8 August 2026, read 9 August 2026
  • Around July 2025, Cloudflare observed crawl-to-referral ratios ranging from 118:1 up to nearly 50,000:1 across leading AI companies.Source: Cloudflare, “Unmasking the crawls with Attribution Business Insights,” 1 July 2026
  • 52% of crawler requests were for AI training as of June 2026, up from 22% in spring 2025.Source: Cloudflare, 1 July 2026
  • More than one-third of crawler activity comes from mixed-use bots that do not distinguish crawling for search from crawling for training, leaving publishers unable to permit one and refuse the other.Source: Cloudflare, 1 July 2026
  • Google’s combined discovery-and-AI crawler gives it access to roughly twice as much information as leading AI companies that separate the two functions.Source: Cloudflare, 1 July 2026
  • More than 50% of traffic on the internet is now non-human.Source: Cloudflare, 1 July 2026
💡
Two figures that look contradictory and are not

AI platforms account for under 1% of publisher referrals, yet crawl-to-refer ratios run into the hundreds and thousands. These measure opposite ends of the same exchange. The crawl ratio counts pages requested per visitor returned by a given operator; the sub-1% figure counts what share of a publisher’s total incoming traffic arrives from chatbots. Both point the same way: content is being consumed at scale and traffic is not coming back in proportion. Quoting them together is more informative than either alone, provided the difference in what they measure is stated. For a fuller breakdown of the crawling side, see our bot traffic statistics.

Cite this page

Paid Hosting. AI Search Traffic Loss Statistics 2026. August 2026.
https://www.paidhosting.com/ai-search-traffic-statistics/

Figures, charts and calculations on this page may be reproduced with attribution and a link to this page. Where a statistic is attributed to a third party, please cite that source directly. Press and data enquiries: [email protected]

Sources & Methodology

Every statistic on this page was read directly from the source named beside it and carries that source’s publication date. Figures marked as Paid Hosting calculations are arithmetic performed on published source data, with the inputs stated so the working can be checked. Where a figure reached us through an intermediary we say so and read the intermediary directly: Chartbeat’s data was provided exclusively to Axios, and we verified the figures against the Axios article itself rather than secondary coverage of it. Claims that appeared only as third-party studies cited inside another publisher’s analysis have been excluded from this page rather than repeated at second hand. Live measurements carry the window they cover and the date they were read. Where sources disagree or measure different things, we explain the difference rather than choosing one. This page is updated as new data is published.

  • Similarweb — “Zero-Click Marketing: The SEO Strategy For A World Where Most Searches Skip Your Site,” 10 June 2026. Clickstream panel data, shared with Rand Fishkin of SparkToro for his June 2026 zero-click study.
  • Similarweb — 2025 Gen AI Landscape report (US desktop, September 2025) and 2026 AI Brand Visibility Report (January 2026 US consumer panel).
  • Chartbeat — publisher traffic data aggregated from thousands of client websites globally, provided exclusively to Axios. Sara Fischer, “Exclusive: Small publishers hit hardest by search traffic declines,” Axios, 17 March 2026. Read at the Axios original rather than via secondary coverage.
  • Reuters Institute for the Study of Journalism, University of Oxford — Digital News Report 2026, 16 June 2026. Fifteenth annual edition; total sample 97,520 across 48 markets, approximately 2,000 per market.
  • Pew Research Center — “Google users are less likely to click on links when an AI summary appears in the results,” 22 July 2025 (browsing data from 900 U.S. adults who consented to share activity, March 2025).
  • Cloudflare — “Content Independence Day, one year on,” 1 July 2026, and “Unmasking the crawls with Attribution Business Insights,” 1 July 2026.
  • Cloudflare Radar — live crawl-to-refer ratios, worldwide, read 9 August 2026 covering 2–8 August 2026.

A note on sources: Similarweb, Chartbeat and Cloudflare each measure only the traffic crossing their own networks or panels, and each sells products related to what they measure. Similarweb in particular sells AI search visibility tools, and its analysis argues for a strategic shift its products support. We cite them because clickstream panels and publisher analytics are the only instruments that measure this behaviour at scale. Two non-commercial sources provide the independent check: Pew Research Center, which observed real browsing rather than surveying, and the Reuters Institute at Oxford, whose Digital News Report surveys roughly 97,520 people across 48 markets and publishes its methodology in full. Where the commercial and non-commercial figures can be compared — the scale of the search referral decline — they agree to within a percentage point. Where they diverge — self-reported versus observed click-through — we say so rather than choosing the more convenient number.