TL;DR
Between January 2024 and February 2026 small publishers lost about sixty percent of their search traffic, mid-sized ones about forty-seven, and large ones about twenty-two. In the same period the New York Times grew its revenue by eleven percent. That gradient is the whole story, and it means Google's claim that clicks are stable and the publishers' claim that they are collapsing can both be true. What is dying is not journalism. It is the middle and the bottom of it, and the history of how commons collapse says that is the normal shape.
Read the gradient, not the average
Chartbeat measured search traffic across thousands of client sites over two years, and Axios published the numbers in March.
Sites with a thousand to ten thousand daily page views: down 60 percent. Ten thousand to a hundred thousand: down 47 percent. Above a hundred thousand: down 22 percent.
Now put the New York Times next to that. Second quarter 2026, filed with the SEC on 5 August: revenue 762,5 million dollars, up 11,2 percent. Digital-only subscribers 12,80 million, up about a million and a half in a year.
Google said in August 2025 that the total volume of organic clicks to sites had stayed relatively stable year over year, and that click quality had gone up. It published no dataset and no method, which is a reason to treat the claim carefully and not a reason to assume it is false. Because here is the uncomfortable part: it can be true. An aggregate can hold steady while its distribution collapses inward. Stable in total and minus sixty at the bottom are the same sentence, told by people with different exposure.
The mechanism is measured, and it is small
Pew put a behavioural panel on this rather than a clickstream estimate. Nine hundred US adults, 68.879 real Google searches in March 2025.
When an AI summary appeared, users clicked a traditional link in 8 percent of visits. When it did not, 15 percent. Links inside the summary itself got clicked in 1 percent of visits.
That is roughly halving the click, not abolishing it. A brand people search for by name survives halving. A site that existed because it ranked third for a question does not.
Meanwhile the replacement channel is not arriving. Chartbeat found referrals from ChatGPT still under one percent of total page views, even after growing two hundred percent. Two hundred percent of almost nothing.
And the licensing market, which is supposed to be the new revenue, is a product for large accounts. News Corp's deal with OpenAI was reported at over 250 million dollars across five years, Amazon's with the Times at twenty to twenty five million a year. Both figures come from unnamed sources and neither has been confirmed by the parties. Whatever the real numbers, nobody is negotiating a licence for a site with four thousand readers a day.
Newfoundland, 1992
The northern cod fishery is the closest structural match I know, and not for the reason people usually reach for it.
Offshore catch peaked at 810.000 tonnes in 1968. Spawning biomass fell about 93 percent in thirty years, from 1,6 million tonnes in 1962 to somewhere between 72.000 and 110.000 by 1992. Canada declared a moratorium in July 1992 and roughly 30.000 people in Newfoundland and Labrador lost their work.
The detail that matters is not the collapse. It is this:
Commercial catch rates stayed deceptively high through the 1980s, because advancing technology could still find and take the fish however few of them there were.
The instrument masked the depletion. Sonar found the last dense schools, so the number that everyone watched kept looking fine while the thing it measured was disappearing.
And the damage was regressive. The moratorium destroyed the inshore small-boat fishermen. The factory fleet converted, or moved, or was already somewhere else. A commons does not collapse onto everyone equally. It collapses onto whoever had nothing but the commons.
The moratorium was lifted in 2024, thirty two years later. The biomass is around twelve percent of its 1960s level. Stopping the extraction did not bring the resource back.
Why this matters for your business
If you have a content strategy that depends on ranking, you are in the part of the distribution that is losing sixty percent, whatever your traffic looked like in 2023.
The publishers that grew did not out-optimise anyone. They stopped depending on search before they had to. Direct relationship, paid or not: an email list, an app, a reason to type your name into the bar rather than a question. That is the only asset in this whole picture that did not decline.
Build the thing people come back to on purpose. The rest is fishing a stock somebody else is still measuring for you.