Street-level language for Paris AI answers

Standing near République, I once watched three people describe the same short walk in three different cities: “toward Canal Saint-Martin,” “up by Oberkampf,” and “near the 10th, sort of.” The map had not moved. The language had. That is the Paris problem I keep seeing in AI answers. In a composite prompt check, a quiet cafe near the canal turns into a tourist Marais suggestion. In another, a local salon in the 15th is treated like a broad central-Paris choice. A bakery can be correct on the map and still vanish when its own pages leave it floating between Jourdain, Pyrénées, Belleville, and plain “Paris.”

I grew up just outside the périphérique, close enough to learn Paris by metro exits before I learned it by districts. Parisians often speak in shortcuts: “côté Jules Joffrin,” “near the market street,” “past the bridge,” “not the tourist Saint-Germain.” Visitors lean on landmarks and arrondissement numbers. Locals may skip the arrondissement and name the slope, the station, the square, or the side of a boulevard. Those differences matter because AI systems do not experience the walk. They inherit the words left behind on homepages, menus, booking pages, directories, reviews, and fragments of bilingual copy.

Before this work, I handled neighbourhood copy audits, bilingual menu and service-page cleanup, local directory evidence review, intake-form diagnosis for appointment businesses, and small shopfront visibility checks. I am strongest where language meets place: when a business near an arrondissement border needs protection from the larger label, or when an English query and a French query produce two versions of the same place. I do not try to make every Paris business sound central, fashionable, or visitor-facing. A local gym by a metro exit, a lunch bar serving nearby offices, a wellness practitioner known to residents of one quartier: each needs its real local shape, written plainly enough that AI can keep it.