
Buyers research before they spend. In roofing they ask which brand is better, what steel thickness they need, which accessories the system requires and what the whole roof will cost. Every category has its version of those questions.
More of them now get asked to ChatGPT or Gemini, answered in a paragraph, and the buyer goes to whichever source the answer names. If you are not in the answer you are not in the consideration set, and nothing tells you that you were cut.
Being in the answer takes hundreds of pages that answer real buyer questions clearly enough for a machine to use. Most companies cannot staff that. Darleot could not either. They sell roofing systems.
It builds one record of the business first. Recomaze connected to Darleot's site and built their Company AI Brain: what they sell, which brands they carry, what each system is actually for, who buys it. Everything downstream reads from that one record, which is why the pages, the guides and the on-site conversations all say the same thing about the company.
Content Studio writes the catalogue from it. Every product page on Darleot's site is written by Recomaze, along with the buying guides and comparisons that sit around them. No briefs, no agency, no writer.
AI Search runs the loop on autopilot. Recomaze identifies the real buyer questions in the category, scores where Darleot stands on each one, writes what is missing on a schedule Darleot set and switched on, then measures what each published piece moved. The questions Darleot misses become the next things it writes. Nobody at Darleot runs this.
The Sales agent converts the visitors already there. Same record, so it knows the catalogue and the systems. Darleot made it theirs: their colours, a name they chose, an avatar they created. A first-time builder who arrives saying they do not know where to start gets walked to the specific ridge, sealing and drainage components their roof needs.
One record, one subscription, work coming out of three surfaces.
The AI engines now cite pages Recomaze wrote. On product questions about matte-finish metal tiles, ChatGPT names Darleot in first position and pulls its cited sources from Darleot's product pages, the ones Content Studio generated. The content is not sitting on the site waiting to be found. It is what the answer is built from.
Gemini describes Darleot correctly, because the record is correct. On calculator questions Gemini names Darleot in first position with positive sentiment, citing Darleot's own domain, and describes the tool accurately: compatible with the major roofing brands, able to estimate single-slope, two-slope and four-slope roofs, L-shaped roofs and dormers. That precision came from the Company AI Brain. An engine can only describe you as well as your record lets it.
Google's AI answers point to the calculator. The tool Darleot built to help buyers plan a roof is what Google's AI answer now sends people to.
The loop keeps running. Every week Recomaze reads which buyer questions Darleot appeared in and which it missed, and the misses set the next brief. That is the part that compounds, and it is the part nobody at Darleot has to think about.
Nothing here rests on our word. Ask ChatGPT or Gemini the questions buyers ask in this category and read the sources. Inside Recomaze the same evidence sits at two levels: per article, what each piece moved, and per buyer question, which engine, which position, which sources it cited.
See where you stand. Start free, connect your site, and Recomaze will show you which buyer questions you already appear in and which ones you are missing.