Small Publishers Lost 60% of Google Referrals While Large Ones Lost 22%

The headline number everyone repeats is that Google search traffic to publishers fell by about a third over the past year. It’s the wrong number to repeat, because the average hides the only fact that matters.

Chartbeat’s breakdown splits the decline by publisher size. Small publishers lost roughly sixty percent of their search referrals. Medium ones lost forty-seven percent. Large ones lost twenty-two percent. Same algorithm, same year, three completely different outcomes.

That spread is not a rounding artifact. A small site losing sixty percent of search is a site losing its business. A large site losing twenty-two percent is a site having a bad year and cutting a budget. One of those is survivable.

Individual results are stranger still

Politico’s organic Google traffic fell around twenty-three percent year over year. CNN lost about a quarter. USA Today’s national edition lost close to half. Business Insider lost more than eighty-five percent.

Business Insider is the number that should worry every publisher who thinks scale is protection. That’s a large, well-resourced, professionally optimised newsroom losing nearly all of its search visibility inside a year. Whatever protected CNN did not protect them, and the difference has less to do with size than with what kind of query the site was winning.

Sites built on informational queries got hit hardest, because informational queries are exactly what an AI answer handles well. If your traffic came from people wanting to know a fact, a definition, an explainer, a how-many or a what-time, that traffic was always vulnerable. It just wasn’t obviously vulnerable while it was working.

The mechanism is boring and total

Every Google search now returns a generated answer as the primary result. Zero-click searches, where the user never leaves for a third-party site, run at roughly sixty percent of all queries. There’s no penalty being applied, no update to recover from, no technical fix on the publisher side. The interface changed, and the interface was the business.

Reuters Institute’s survey of news executives puts the forward expectation at a forty-three percent drop over three years, with a fifth of respondents expecting to lose more than three quarters of their search traffic. Note that these are the people running the newsrooms, forecasting their own decline. Optimism bias usually inflates numbers like these upward. It didn’t here.

What the spread actually produces

If you compress the losses into a single average, you get a story about AI versus journalism. If you look at the spread, you get a different story, and it’s about consolidation.

The long tail of the web is being cleared. Independent blogs, niche trade sites, regional outlets, single-topic publications, the specialist coverage that never had a direct audience and never needed one because search delivered readers reliably for fifteen years. That layer is going, quickly, and it’s going without any of the coverage that a comparable wave of newspaper closures would get.

What survives is what was already large enough to have a direct relationship with readers, plus whatever the AI companies choose to license. That’s a narrower information supply, sourced from fewer organisations, summarised by systems trained on the layer that’s disappearing.

The awkward part is the feedback loop. The answer engines are good because the open web was rich. The open web was rich because small publishers could fund themselves on search traffic. Remove the funding and the source thins out, and the thinning shows up in the answers a few years later, by which point there’s nothing to rebuild from.

Nobody chose this. It’s just what happens when the interface changes and the incentives don’t.