2026 Barometer of French SMEs' Readiness for AI Search (GEO/AEO)

This dataset measures the technical readiness of French small-and-medium-business (SMB) websites for AI answer engines (ChatGPT, Perplexity, Gemini, Copilot) — i.e. not "which brands the AIs cite", but "which sites the AIs are technically able to read, understand and cite". It covers a population no existing barometer measures: independent French SMBs that already rank in classic search. Sample: 474 websites analysed (509 collected, 35 unreachable), across 10 business sectors x 15 cities, drawn from real local Google SERPs (top-10 organic, collected via Brightdata on 2026-07-13). Directories, marketplaces, booking platforms, national franchises and public institutions are excluded; deduplicated by domain. Headline results: mean readiness score 48/100 (median 49); 80.8% below 60/100; only 1.3% score >=80. Agentic navigation 1.2/3 on average. 19.6% have an llms.txt file (13.3% spec-compliant); only 1.3% block AI bots in robots.txt while 85.9% leave access technically open. The gap is not access but exploitability: weak answer structure (17.4/35), weak trust/E-E-A-T signals (15.6/35), author absent on 82.1% of homepages. Method: each homepage is scored by Les Créavores' static AEO engine (HTML + robots.txt + llms.txt checks, no LLM call): score/100 = AI crawlability (30) + answer structure (35) + trust/E-E-A-T (35); agentic navigation/3 = accessibility tree + spec-compliant llms.txt + open agent access. Homepage only. Published data is aggregated by sector and city; no domain name is disclosed, no company named or ranked individually. Companion study (FR): https://lescreavores.fr/barometre-referencement-ia-tpe-pme/ Mirror DOIs: https://doi.org/10.5281/zenodo.21381109 (Zenodo) — https://doi.org/10.6084/m9.figshare.32996552 (Figshare) — https://doi.org/10.7910/DVN/IIB2AW (Harvard Dataverse)