# v1.1 — editorial correction (2026-11-09)

- Freeze the historical CSV; no source verdicts or domains changed.
- Separate 12 AI-related tokens from two ordinary search baselines.
- Recompute every number; correct rank buckets, named-group counts, redirect rounding and median size.
- Compare restriction conditional on heuristic llms detection, rather than reversing the probability.
- Replace citation-outcome and operator-motive claims with root-path declarations.
- Correct figure labels, bilingual output, download and Dataset URLs.
- Disclose missing llms bodies, unresolved retry/error cases, incomplete source-list provenance and alias bias.
- This is not the later method-validated or re-collected dataset.