Fontaine appears in roughly a quarter of the AI answers a customer in Lyon 6e sees. The gain since May is real — it sits outside the noise band.
Ranges overlap month to month. We only call something a win when it clears the band — ChatGPT +3 does, AI Overviews −1 doesn't.
Google surfaces (AI Overviews, AI Mode) lag the rest — and they lean hardest on listings. See sources, p. 04.
5 of the 7 most-cited sources are editable listings or review profiles. Fixing presence there is routine agency work — it just now moves AI answers too.
Cited in 64% of answers; Fontaine absent. Highest-leverage fix on the board.
210 reviews vs top rival's 480. Engines echo review language — set up a post-job ask flow for techs.
Own site already cited in 22% of answers. District-level service pages get cited ~3× more than city-generic ones.
Local press citations at 9% and rising across engines. One placement compounds across every engine.
Schema markup hacks, llms.txt files, bulk backlink packages — across 1,800 sampled answers we measure no effect from any of them. If a vendor promises AI placement, ask for their sampling method.
A mention rate of 12% ±5 means: across 30 sampled runs, Fontaine appeared in 12%, and we're 90% confident the true rate sits between 7% and 17%. Month-over-month changes inside that band are reported as noise — even when they'd look good on a chart.
Conversions, calls, or bookings — pair this report with your call-tracking. We also don't measure prompts nobody asks; the pack is built from real query patterns for the vertical + city.
jun 14 — AI Overviews layout change in fr results; local-pack citations shifted. Flagged, not treated as client movement. · jun 02 — AI Mode added to the panel (first month; no trend yet).