Most of what is written about AI search is advice: confident, unfalsifiable, and untethered to data. It rhymes with the SEO of 2005 — a folklore economy where the loudest voice won and nobody kept score.
SEO matured when it started measuring: controlled tests, holdout groups, reproducible results. GEO is at the pre-measurement stage now, and the same maturation is available to whoever does the unglamorous work first.
This archive picks the measurement side, explicitly. Controlled corpora. Reproducible runs. Published method. Datasets on request. If a claim here cannot be tested, it does not belong here. That is the entire editorial policy.
