Skip to main content
Answer Engineered

Google's share of publisher traffic: 23%, 7.3% or 5%, depending on who counts

Six dated releases on publisher traffic from Google, read for what each counted. Three shares, four declines, and one 25% headline that was a range edge.

7.3%vs0.02%

Share of publisher referrals in the Chartbeat data behind the Reuters Institute 2026 trends report, as Digital Content Next summarised it (5 October 2026): Google Search 7.3%, Google Discover 13%, ChatGPT 0.02% and Perplexity 0.002%. A share of referrals across more than 2,500 sites, not of pageviews or of inbound traffic, which is why other releases print 5% and roughly 23% for Google

Six releases, and what each one counted

ReleaseDateWhat it countedHeadline figure
Similarweb, news publishers reportJune 2025organic visits to news sites, Similarweb panel, peak mid-2024 to May 2025down 26%
Google, Search blog6 Aug 2025total organic clicks to all websites, internal data, no figures”relatively stable”
Digital Content Next, member survey14 Aug 2025median change in Google Search referrals, 19 member companies, eight weeksdown 10%
Reuters Institute, Chartbeat data12 Jan 2026Google organic search traffic to over 2,500 sites, Nov 2024 to Nov 2025down 33%
Chartbeat, 2026 Publisher PlaybookOct 2026Google Search, year on year to July 2026, Chartbeat customersdown 40.2%
Digital Content Next, CEO post27 Aug 2026organic search as a share of inbound traffic, unnamed set of membersroughly 23%

Similarweb. The landing page of its generative AI and publishers report says “Organic traffic to news sites has dropped 26% since Google launched AI Overviews”. The full report sits behind a download form. TechCrunch’s coverage gives the absolute numbers behind it: organic traffic fell “from over 2.3 billion visits at its peak in mid-2024 to now under 1.7 billion”. By our arithmetic, 1.7 over 2.3 is a fall of about 26%, so the 26% runs from a peak, not year on year. The same article says the share of news searches that end without a click to a news site “has grown from 56% to nearly 69% as of May 2025”. That is news searches, not all searches. Similarweb is a vendor reporting its own panel, and no sample size is published on the page.

Google. Its 6 August 2025 post says: “Overall, total organic click volume from Google Search to websites has been relatively stable year-over-year.” The same post concedes the other half: “user trends are shifting traffic to different sites, resulting in decreased traffic to some sites and increased traffic to others.” No number, no base, no window. It is a statement about all websites in aggregate, so it does not contradict a decline for publishers, and it cannot be tested either.

Digital Content Next. The August 2025 post reports “median year-over-year referral traffic from Google Search to premium publishers down 10% over just eight weeks”. The data “spans 19 companies”, the window is “Over eight weeks in May and June 2025”, news brands fell 7% and non-news brands 14%, and declines came “most in the -1% to -25% range”. It is a median of company-level changes from a trade body that argues for regulation, which the post does not hide.

Reuters Institute. The 2026 trends report says Chartbeat data “show that Google traffic from organic search to over 2,500 sites was down by a third (33%) globally between Nov 2024 and Nov 2025 and by 38% in the United States, but it is not clear how much of this is down to AI overviews.” The second half of that sentence is the one that rarely gets quoted.

Chartbeat. Its 2026 Publisher Playbook says Google Search was “roughly 9% of pageviews” in July 2024 and 5% by July 2026. The playbook PDF puts Google Search down 21.9% from July 2024 to July 2025 and down 40.2% from July 2025 to July 2026, and gives the biggest shares of pageviews to internal traffic (40.4%) and direct traffic (15.4%). We went through its windows, and why its figures do not line up with each other, in the Pew and Chartbeat entry.

Digital Content Next again. A 27 August 2026 post by the group’s CEO says organic search “still represents roughly 23% of inbound traffic across a representative set of DCN members, down from 25% a year earlier”. The members are not named, the count is not given, and “inbound traffic” is not defined. It is a figure in an argument, not a release with a method.

Which numbers get put together that should not be

The 25% that was a range edge. The day after the DCN survey, Digiday ran it as “Google AI Overviews linked to 25% drop in publisher referral traffic”. DCN’s median was 10%. The 25% is the far end of the range most members fell in. And DCN measured Google Search referrals, not exposure to AI Overviews, so “linked to AI Overviews” is DCN’s reading, not its measurement.

Ten percent and 33% are not a range. DCN’s own October 2026 post quotes the Reuters figure next to its survey and adds, fairly: “The methodologies and measurements differ.” One is a median across 19 companies over eight weeks. The other is an aggregate across more than 2,500 sites over a year. They cannot be averaged, and “publishers lost between 10% and 33%” is a sentence neither study supports.

Three shares, three denominators. 23% of inbound traffic, 7.3% of referrals, 5% of pageviews. Pageviews include direct visits and clicks inside the site, which together make up 55.8% of pageviews in Chartbeat’s playbook by our addition, so a channel’s share of pageviews will come out smaller than its share of referrals. A publisher comparing its own Google share with one of these should first check which of the three it computed.

Peak-to-now is not year on year. Similarweb’s 26% runs from a mid-2024 peak. Reuters’ 33% runs November to November. Chartbeat’s 40.2% runs July to July. Put side by side without the windows, they read like an acceleration. They may be one, but these three releases cannot show it.

What AI sends instead

Some of this volume is supposed to come back through chatbots. On the evidence so far, it does not, yet. In the same Chartbeat dataset, as DCN summarised it, “Google Search accounted for 7.3% of publisher referrals and Google Discover another 13%. ChatGPT accounted for just 0.02% and Perplexity 0.002%.” The figures come from the referral-share chart in the Reuters Institute report.

Source of referralsShare of publisher referrals
Google Discover13%
Google Search7.3%
ChatGPT0.02%
Perplexity0.002%

The report’s own prose says “Google delivers 500 times as many referrals as ChatGPT from search alone”. By our arithmetic the printed shares give 7.3 over 0.02, or 365 times. The chart values are rounded, so the two need not disagree, but at the printed precision the 500 cannot be read off the chart. Chartbeat’s playbook puts it on a pageview base instead: “total AI referral share remains under 0.01% of pageviews.”

What to measure on your own site instead

Borrowing an industry figure for a single site fails for the same reason the figures disagree with each other. Four checks, each on your own analytics:

  1. Pick one denominator and keep it. Referrals, sessions or pageviews, written down next to every share you report.
  2. Compare like windows. Year on year for the same weeks, not peak to now.
  3. Split Google Search from Google Discover and Google News. The Reuters chart counts them separately, and its Google Search category excludes Google News.
  4. Report the median of your sections or properties as well as the total, the way DCN did, so one large section cannot carry the whole trend.

None of this says how much of a decline AI answers caused. None of the six releases claims to know that either, and the Reuters report says outright that it is not clear.

Methodology limits

All six releases were opened on 4 November 2026. Similarweb’s full report is gated, so its 56% to 69% figure comes from TechCrunch’s reporting of it. The Reuters referral shares are read from the report’s chart, and the four we use match the figures DCN printed from the same dataset; we do not use the chart’s other channels. Chartbeat publishes no site count for the playbook. DCN’s 23% has no published method. Google’s statement has no data at all. Every ratio and subtraction above that is not in a source is our arithmetic and is labelled as such.

Frequently asked questions

How much traffic are publishers losing to AI?

No single number answers it, because each release counts something different. Digital Content Next reported a median 10% fall in Google Search referrals across 19 member companies over eight weeks in May and June 2025, while Chartbeat data in a Reuters Institute report show Google organic search traffic to over 2,500 sites down 33% globally between November 2024 and November 2025. The Reuters report says it is not clear how much of that is down to AI Overviews.

Where does the 25% drop headline come from?

From Digital Content Next's May to June 2025 member survey, where most members' Google Search referrals fell by between 1% and 25%. The median was down 10%. Digiday headlined the 25% edge of that range.

It depends on the denominator. Digital Content Next puts organic search at roughly 23% of inbound traffic for a representative set of its members, the Reuters Institute's Chartbeat data put Google Search at 7.3% of publisher referrals, and Chartbeat's 2026 playbook puts it at roughly 5% of pageviews in July 2026.

Does Google say publisher traffic is falling?

Google said on 6 August 2025 that total organic click volume from Google Search to websites has been relatively stable year-over-year, with traffic shifting so that some sites lose and others gain. It published no figures, and the statement covers all websites, not publishers.

How much traffic do AI chatbots send publishers?

In the Chartbeat data published by the Reuters Institute, ChatGPT accounted for 0.02% of publisher referrals and Perplexity 0.002%. Chartbeat's 2026 playbook says total AI referral share remains under 0.01% of pageviews.