I read rankings as arranged proof
My work sits between search auditing, editorial diagnosis, and old-fashioned comparison reading. I look at why a French business is named first, fifth, or not at all when AI systems answer “best,” “top,” and “compare” prompts. The useful question is usually narrow: which public proof helped one firm keep a higher seat?
About
Before rewriting anything, I want to know whether the business failed to enter the answer or failed to deserve its seat.
Two rival hotels, one wet café table in Nantes, and a notebook with arrows running from room descriptions to review snippets to local guide blurbs. That was the first ranking map I kept instead of throwing away. The AI answer had placed the louder hotel above the quieter, better-matched one. The gap was not mysterious. One business had public wording that could be picked up, classified, and compared. The other had praise, but praise in soft fog.
I am from western France, and I have spent seventeen years around search audits, editorial strategy, local-business copy diagnosis, comparison-page analysis, and quiet advisory work for service companies. The thread is simple enough: I read what a business says about itself, what other sources repeat, what gets stale, what gets translated badly, and what a ranking answer can use when it needs to choose an order. I still draw ranking maps by hand because the pencil slows me down. It keeps inclusion signals on one side and ordering signals on the other. Those are different jobs. A business can be eligible for a list and still lose the first three seats because its proof is vague.
Now I work mostly with French businesses that depend on being compared: hotels, clinics, schools, agencies, restaurants, manufacturers, and specialist service firms. I compare French and English prompts because the two trails often disagree. A business may look strong in French, then become flat in English, stripped of its local angle and placed behind a rival with cleaner evidence. Review score is one signal among others. Reviews matter, but so do category labels, fresh third-party mentions, service-page sentences, regional detail, and proof that a system can quote without guessing. My stance is plain: AI best-of answers are arranged evidence. If the public trail is mushy, the ranking will usually be mushy too.
If the answer feels wrong, the evidence trail probably explains why.
Send the prompt, the rival, and the seat you expected to hold. I will tell you whether the issue is visibility, ordering, or both.
Ask for a reading