MOUNTAIN VIEW, 22 AUG 2026 — Organic Google traffic to Business Insider fell by more than 85 per cent between June 2025 and June 2026. USA Today's national site lost close to half. CNN fell about a quarter and Politico 23 per cent.
An antitrust filing puts the click-through decline associated with AI Overviews at 58 per cent. Reddit, USA Today and Reuters are among the publishers weighing whether to block Google's crawler entirely.
The numbers
Google announced five changes to AI Overviews and AI Mode in May intended to send more traffic to publishers, including a further-exploration links section, subscription labels and inline link context. In January the United Kingdom's Competition and Markets Authority published proposed conduct requirements including publisher controls, allowing publishers to opt out of powering AI features and of training use outside Google search.
The spread between publishers is the informative part
A decline of 23 per cent and a decline of 85 per cent are not the same phenomenon, and treating them as one loses the analysis.
The publishers hit hardest, like Business Insider, relied on traffic from queries an AI summary can answer completely. Once a definition, a how-to, or a product basic appears on the results page, the click has no purpose.
Publishers with smaller declines offer things a summary cannot substitute — original reporting people seek by name, coverage where the source is the story, and audiences arriving out of habit rather than through search. Politico's readership is substantially professional and direct.
Search referral is not collapsing across the board. Instead, content that just answers a question is losing its distribution channel, while content that people read because of who published it is holding up. Those were always different businesses, and search traffic disguised the difference for twenty years.
Blocking the crawler is a harder decision than it sounds
The retaliatory option receives more attention than it deserves, because the asymmetry runs the wrong way.
A publisher blocking Google's crawler forgoes whatever referral remains, immediately and completely, in exchange for denying Google content it may substitute for. If traffic is already down 85 per cent, the calculation shifts. The publisher has little referral left to lose, but their content is still being used. This explains why the publishers considering the move are the ones with the steepest declines.
What makes it harder is that the crawl is not one thing. The crawler that indexes for search, the retrieval that grounds an AI answer and the collection that trains a model have historically been difficult to separate through robots directives, which is exactly what the CMA's proposed publisher controls address. A publisher wanting to be in the index but not in the summary has not reliably had a way to say so.
Until that separation is enforceable, blocking is an all-or-nothing lever, and few publishers can afford to pull it while any referral remains.
What the five May changes did and did not address
Further-exploration links, subscription labels and inline link context all address link visibility. None addresses the underlying dynamic.
A more prominent link to an article a reader no longer needs does not help. Improving link placement might capture marginal clicks from readers who want the source, but it will not bring back those whose question was already answered. The 58 per cent click-through decline represents people who got what they came for on the results page.
Subscription labels are the most interesting of the three because they are a concession. The label warns readers that a link leads to a paywall, effectively signalling them to avoid subscription publishers — an odd remedy to offer.
We are not a neutral observer here
RECATOOLS is a publisher, and disclosure is cheaper than pretending otherwise.
Our own indexation position is worse than the figures above describe, for a different reason. Google has never crawled roughly 3,232 of this site's 3,930 URLs, and the monthly crawl rate has been falling — 300 URLs in June, 236 in July, 162 in August by most-recent-crawl month. The problem here is not summaries replacing clicks on pages that rank. It is pages that were never fetched.
Those are different failure modes with the same consequence, and they interact badly. A publisher struggling for crawl budget has less indexed content competing for the referral that AI summaries are already reducing.
What we learned trying to fix this is worth passing on. On-site remedies did not work. Site-wide navigation anchors on every page did not get three of our hubs crawled at all. If a site-wide link cannot get a hub fetched, no internal linking scheme, sitemap change or canonical strategy will move article pages. That is a discouraging finding and it is better published than kept.
Why this matters for regional publishers specifically
The declines above are measured on large English-language publishers with brand recognition, direct traffic and subscription businesses. Publishers across ASEAN generally have less of all three.
A regional publisher whose audience arrives predominantly through search, without a subscription base or app habit to fall back on, has fewer buffers than the outlets producing these figures. The mitigation that works for a national title — being the destination people name — is expensive and slow to build.
The second regional wrinkle is language. AI summarisation quality varies by language, and a summary that answers a query well in English may answer it poorly in Bahasa Indonesia, Thai or Vietnamese. That is a temporary reprieve rather than a moat, and it will narrow.
The honest conclusion is that a publishing business dependent on search to answer questions has a structural problem no on-site SEO can fix. The only levers left are off-site: building a direct audience, getting referral from other platforms, and producing coverage nobody else has.
What remains unconfirmed
The traffic figures are drawn from reporting rather than from publisher disclosures, and the measurement methodology and data source are not described in the material reviewed. The 58 per cent click-through figure originates in an antitrust filing, which is an advocacy document produced by a party with an interest in the number.
Google has not confirmed these declines and disputes this framing. It also is not clear whether any named publisher has actually blocked the crawler, or just considered it. The status of the CMA's proposed publisher controls remains unstated, and no comparable figures for non-English or regional publishers are available.
What to watch for
The first thing to watch is whether a major publisher actually blocks. The threat has been made repeatedly; the first substantial outlet to follow through will produce the only real evidence about what blocking costs.
The second is whether the CMA controls become enforceable. Separable opt-outs for indexing, grounding and training would change the negotiation from all-or-nothing to something a publisher can actually price.
The third is any Google disclosure of referral data. Every figure in this story is measured from outside; the party that can settle it has not.
