# J+30 methodology

Version 1.0.0 — 2026-09-26

MIA-FR replays 25 non-branded French prompts across four platforms and three repetitions, for 300 planned runs. Five non-valid runs remain in the plan but are excluded from visibility rates. J0/J+30 differences use a prompt-paired bootstrap.

Quality control covers 30 rows: 29 can be fully recoded and one has an empty raw response. This is a separate documentary pass performed by Codex, not an independent human review. Two mention disagreements are retained without forced consensus.

SEO/GEO replays 20 queries verbatim on Google France, Google AI Mode and Perplexity. One pass is retained. Incomplete responses are neither imputed nor converted into citation absence.

The instruments remain separate: MIA-FR measures repeated frequency; SEO/GEO describes availability and visible sources across search surfaces. Results are descriptive and do not establish causality. Raw responses and screenshots remain in the local evidence archive; the public distribution contains structured observations, methodology, limitations and checksums.
