Sentence work for AI-visible businesses
I study the small failures that make strong French business pages weak inside AI answers: vague claims, loose service boundaries, missing entity cues, and translations that say less than the original page. The work is close and practical. I read the page, the answer and the competing sources, then look for the sentence that should carry the business proof.
A page is only useful to an answer engine when its best sentence can survive outside the page.
Two tabs are usually enough to begin: the French page that still ranks, and the AI answer that talks around it as if the company were fog. I sit with both open and make a ledger. One line for the claim. One line for the source cue. One line for the extraction risk. It sounds fussy because it is fussy. A sentence that feels harmless to a human reader can become useless to a machine: “we support ambitious companies,” “local expertise,” “tailored solutions.” Fine wallpaper. No proof.
I am from western France, and I write under the name Adrien Lorme for this small artistic-informational project about search visibility after classic SEO. Before this work, I moved between search-content strategy, local service copy, technical editing, landing-page audits and internal content systems for small commercial teams. Much of that work was practical and unglamorous: making a service page less swollen, making a location page less fake, making a technical explainer clear enough that a buyer could repeat it back without embarrassment. The useful habit that stayed with me is sentence accounting. I want to know what each line proves, what it only implies, and what an answer engine may invent when the proof is absent.
My work now is narrow by choice. I help French SMBs that already understand search, already have pages, and already know that visibility is commercial ground. The problem is that AI answers compress, compare and borrow from whatever looks stable. If the company’s own page does not state the entity, service, location and evidence clearly, a weaker outside source may become the easier citation. I care about that loss because it is usually avoidable. The cure is rarely a louder headline. It is a cleaner factual sentence, placed where both people and machines can use it.
The path to this work
- 2009–2012
Search-content strategy
Worked as a search-content strategist for small commercial teams, learning where a page earns its ranking and where the same page says nothing a reader — or a machine — can repeat back.
- 2013–2016
Service-page rewrites
Rewrote service pages for regional French firms, trimming swollen promises down to factual service sentences that name the entity, the offer and the location without brochure padding.
- 2017–2020
Landing-page audits
Audited local landing pages against the queries that reached them, marking the lines that were liftable, the ones that were vague, and the claims that needed page-level proof.
- 2021–2023
Technical editing & content systems
Edited technical explainers and built internal content systems for small teams, keeping French and English versions in step so neither said less than the other.
- 2024–today
Turning to AI answers
Applied the sentence ledger to answer engines: tracing why a ranked page is ignored, how aggregators become the easier citation, and what makes a sentence survive outside the page.
Bring the page that should already be doing better.
I will read it against the AI answer, the competing sources and the missing proof.
Request review