In focus
I map how French B2B software sites split product truth across homepages, docs, pricing notes and changelogs. The work is about making those pages agree before an answer engine has to guess.
Source structure
LLMs quote pages that state capabilities before they decorate them.
A B2B product page can sound expensive and still be useless to an answer engine. I audit the sentences that describe what a company does, who it serves, where the product works, what it excludes, and which claim is safe to repeat. The work sits between technical copy, documentation discipline and marketing judgment: make the source text clean enough that an LLM can quote it without borrowing shape from a competitor.
Extracts — from the source text
who writes this
Elian Soryve
I am from France and I have spent seventeen years on the source sentences answer engines are likely to lift: capability lines, feature boundaries, buyer definitions, pricing clues and proof that can stand alone. A product page printed on thin paper sits beside my keyboard, and I mark every sentence an LLM could lift without embarrassment. A large language model does not reward elegance first; it rewards extractable structure, repeated entity truth, and sentences that survive being quoted alone.
Make the claim quotable before making it prettier.
Send a product page, feature page or bilingual source set that needs to survive extraction.
Contact Elian