The State of Organic Search Traffic for Publishers, 2026

For twenty years, the publisher’s relationship with Google was simple in one respect: a page that ranked got clicked. The ranking was the product. In 2025 that stopped being reliably true, and 2026 is the first year we can measure how much has changed with more than anecdotes.

This report assembles the credible measurements published in the last fourteen months — Pew Research Center’s behavioral panel, two Ahrefs click-through studies, Similarweb’s publisher data as reported by Press Gazette, Chartbeat’s referral data as published by the Reuters Institute, and the Reuters Institute’s own survey of 280 media leaders — into a single picture. Each dataset measures something slightly different, which is why we keep them separate below and say what each one can and cannot tell you. Together they describe one thing: the click is no longer the unit of search visibility.

1. How much less do people click?

Two independent methods arrive at the same answer.

Behavioral panel. Pew Research Center recruited 900 U.S. adults who installed a browsing tracker for the month of March 2025. Across 68,879 unique Google searches, 18% produced an AI summary. When a summary was present, users clicked a traditional result 8% of the time. When it was absent, they clicked 15% of the time. Clicks on links inside the summary happened on 1% of visits. And users were more likely to end their browsing session entirely after a page with an AI summary (26%) than after one without (16%).

Search Console aggregate. Ahrefs approached it from the site side. In April 2025 they compared position-one click-through rates for 300,000 informational keywords — 150,000 that trigger an AI Overview and 150,000 that do not — between March 2024 and March 2025, and attributed a 34.5% CTR reduction to the presence of an AI Overview. In February 2026 they re-ran the analysis on December 2023 versus December 2025 data. The reduction had grown to 58%.

Source: Ahrefs, April 2025 and February 2026 studies of 300,000 keywords using aggregated Google Search Console data. The March 2025 points are from the first study, which used a March 2024 baseline; the December points are from the second.

Notice the second series. Click-through fell for informational queries without an AI Overview too — Google’s other answer-first features, featured snippets, “People also ask,” and the general upward creep of the results page all take a share. The AI Overview is the largest single cause, not the only one.

2. How much traffic has actually left?

Click rates describe behavior per search. Referral data describes what reached the publisher. Three sources, three methods:

• Similarweb (reported by Press Gazette, July 2025): organic search visits to news publishers fell from a peak of more than 2.3 billion to under 1.7 billion in May 2025 — roughly a 26% decline in the year after AI Overviews launched.
• Chartbeat (published in the Reuters Institute’s Trends and Predictions 2026): across 2,576 news sites worldwide, traffic from Google search fell 33% between November 2024 and November 2025; in the United States (797 sites) it fell 38%. Google Discover referrals fell 21% over the same period.
• Reuters Institute survey (280 senior media leaders, 51 countries, fielded November–December 2025): respondents expect search referrals to fall a further 43% over the next three years. About one in five expects a loss of more than 75%.

Source: Reuters Institute, Journalism, Media, and Technology Trends and Predictions 2026 (Chartbeat data); Press Gazette reporting Similarweb, July 2025.

Two things are worth saying about these numbers. First, they are for news publishers specifically — the category most exposed, because news queries are exactly the kind an assistant can summarize in a sentence. Second, they are not the same thing as “search is dying.” Google’s own volumes are not falling. What is falling is the share of searches that end in a visit to a publisher.

3. Zero-click, and the shape of the query

Similarweb’s clearest single statistic: the share of news-related Google searches that ended without any click rose from 56% in May 2024, when AI Overviews launched in the U.S., to almost 69% in May 2025. Per-publisher figures for May 2025 ran to 68.8% (Mail Online), 69.2% (BuzzFeed), and 71.2% (People.com) on keywords where an Overview appeared.

Pew’s data adds the detail that matters for strategy: which searches get an AI answer. Short queries rarely do. Long, conversational, question-shaped queries almost always do.

Source: Pew Research Center, July 2025; 68,879 searches by 900 U.S. adults.

This is the long tail, arriving from the other direction. Specific, natural-language questions are the fastest-growing kind of search, and they are the kind most likely to be answered on the results page. A publisher whose pages address broad topics in general terms is competing exactly where the answer engine is strongest. A publisher whose pages hold the specific fact the question is asking for is competing where the answer engine needs a source.

4. Structural or cyclical?

Publishers have lived through traffic shocks before — Panda, Penguin, the mobile shift, the collapse of Facebook referrals (down 43% over the last two and a half years by Chartbeat’s count) — and some recovered. Three features of this one suggest it will not reverse on its own.

The interface changed, not the algorithm. Prior updates re-sorted the same list of links. AI Overviews replace the list with an answer for a growing share of queries; Ahrefs estimates AI Overviews now appear on roughly 16% of U.S. searches and reached more than 1.5 billion users a month by early 2025. The answer-first layout is a product decision, and Google’s AI Mode extends it.

Ranking and being cited have come apart. In July 2025, Ahrefs found that 76% of pages cited in AI Overviews also ranked in the top ten for the query. By March 2026, across 863,000 keywords and four million cited URLs, that figure was 38%; 31% of citations ranked between 11 and 100, and 31% did not rank in the top 100 at all. A publisher can hold its rankings and still lose its place in the answer. (We look at where the citations went in a separate report.)

Most pages never had traffic to lose. Ahrefs’ study of roughly 14 billion pages found that 96.55% receive no Google traffic at all, and a further 1.94% receive between one and ten visits a month. The organic traffic publishers are losing was always concentrated in a thin slice of their archives — the pages that ranked for real demand. That slice is exactly where AI answers now sit.

5. What is still being clicked and cited

The same datasets that document the decline also describe the exceptions, and the exceptions are consistent.

• Sources with specific, official, or verifiable facts are over-represented in AI answers. Pew found government (.gov) sites made up 6% of the sources cited in AI summaries, three times their 2% share of standard results. News sites held 5% in both.
• Fresh content is favored. Across 16.975 million citations, Ahrefs found the URLs cited by AI assistants averaged 1,064 days old versus 1,432 days for organic results on the same queries — 25.7% fresher. ChatGPT skewed newest (958 days); Google’s AI Overviews were the most conservative.
• Content with statistics and citations is more visible in generative answers. In the peer-reviewed GEO benchmark (KDD 2024; 10,000 queries), adding verifiable statistics to a page raised its visibility in generative-engine responses by 33% on a position-adjusted word-count basis, adding quotations by 41%, and citing sources by 28%. Keyword stuffing reduced it by 9%.
• Pages that add information get chosen. Google’s 2022 patent on “information gain” describes scoring documents by how much they add beyond what a searcher has already seen. An answer engine composing a response has the same need: the page that contributes a fact the model did not already have is the page it must attribute.

Put those four together and a profile emerges. The pages that survive the shift are specific rather than general, current rather than archived, sourced rather than asserted, and contribute something — a number, a comparison, a ranking — that the answer cannot produce on its own.

6. How publishers are responding

The Reuters Institute survey shows where leaders are placing their bets for 2026: paid content (76% call it a primary focus), native and brand advertising (64%), events (54%), and platform licensing (37%). Only 20% expect platform licensing to become a significant source of income; 49% expect a minor contribution and 20% expect nothing.

The survey also recorded what publishers plan to publish less of. On a net basis — the share planning to increase minus the share planning to decrease — service journalism scored −42, general news −38, and evergreen content −32. The logic is understandable: these are the categories AI summarizes most easily. We think the conclusion is wrong, and we make that case in a separate report; the short version is that the evergreen content assistants can replace is thin evergreen, and evergreen content built from data is the kind they have to cite.

Source: Reuters Institute, Journalism, Media, and Technology Trends and Predictions 2026; survey of 280 media leaders in 51 countries.

Methodology and caveats

What each dataset measures. Pew observed real browsing by a probability-based panel of 900 U.S. adults for one month (March 2025); its click rates are per search, not per site, and its sample of AI-summary searches is 12,593. Ahrefs’ CTR figures are aggregated Google Search Console data for 300,000 informational keywords; the April 2025 and February 2026 studies use different baselines (March 2024 and December 2023) and are not strictly a single time series. Similarweb’s publisher figures are modeled estimates of visits, reported by Press Gazette. Chartbeat’s figures cover 2,576 sites on its platform and are reported by the Reuters Institute; the panel skews toward established news organizations. The Reuters Institute survey is a non-probability sample of 280 senior leaders. Ahrefs’ citation, freshness, and zero-traffic studies are drawn from Ahrefs’ own crawl and keyword databases, which over-represent English-language and higher-quality pages relative to the whole web.

Why they are not additive. A 58% CTR decline on Overview-triggering keywords, a 33% referral decline, and a 26% visit decline describe different populations over different periods; none is a subset of another. We present them side by side rather than combining them.

What we did not find. No dataset yet separates the decline attributable to AI Overviews from that attributable to other results-page changes over the same period; Ahrefs’ paired-keyword design is the closest, which is why we lead with it.

What to do about it

1. Stop measuring visibility by rankings alone. Track citations and referrals from AI surfaces (AI Overviews, ChatGPT, Perplexity, Gemini, Copilot) alongside clicks. If a page ranks and is not cited, you are losing the query even though your dashboard says you are winning it.
2. Audit your archive for information gain. For each page type, ask what it contains that an assistant could not generate from general knowledge. Pages that fail that test are the ones whose traffic is disappearing; pages that pass are the ones to invest in.
3. Publish specific, sourced, dated facts — and keep them current. Statistics, comparisons, and rankings with a stated source and a visible date are the properties the citation data rewards. A number that is a release behind is a number that gets skipped.
4. Cover topics completely, not broadly. The queries that trigger AI answers are long and specific. A resource that already answers every entity-, place-, and cohort-level question in a subject is the one an assistant can draw on for almost any phrasing of it.
5. Expect the click to keep shrinking, and value the citation. Being named as the source inside an answer is the new placement. It drives fewer visits per query than a blue link did — and it is the only placement that grows as answers replace lists.

Sources

1. Pew Research Center — Google users are less likely to click on links when an AI summary appears in the results (Chapekis & Lieb, July 22, 2025) — 900 U.S. adults, 68,879 searches, March 2025
2. Ahrefs — AI Overviews Reduce Clicks by 34.5% (Law & Guan, April 17, 2025) — 300,000 keywords, March 2024 vs March 2025; updated February 4, 2026 with December 2023 vs December 2025 data showing a 58% reduction
3. Press Gazette — How Google AI Overviews is fuelling zero-click searches for top publishers (Tobitt, July 7, 2025) — Similarweb data, May 2024–May 2025
4. Reuters Institute for the Study of Journalism — Journalism, Media, and Technology Trends and Predictions 2026 (Newman, January 2026) — Survey of 280 leaders in 51 countries; Chartbeat data for 2,576 sites
5. Ahrefs — Google AI Overview citations and top-10 rankings (Linehan, March 2, 2026) — 863,000 keywords, 4 million cited URLs; compares with the July 2025 study
6. Ahrefs — Do AI assistants prefer to cite fresh content? (Law, July 28, 2025) — 16.975 million citations
7. Ahrefs — AI Overviews Have Doubled (25M AIOs Analyzed) — Prevalence of AI Overviews in U.S. searches
8. Ahrefs — 96.55% of Content Gets No Traffic From Google (Soulo, December 2023) — ~14 billion pages
9. Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024) — GEO-bench, 10,000 queries; visibility changes by optimization method
10. Google patent — Contextual Estimation of Link Information Gain (granted 2022), as analyzed by Search Engine Journal