Which champagne houses dominate ChatGPT — and which are invisible

Dom Pérignon appears in 9 out of 15 benchmark prompts because it benefits from three compounding advantages: an exceptionally rich Wikidata entity (with 40+ properties and sitelinks in 15 languages), decades of coverage in premium media sources (Decanter, Wine Spectator, Wine Advocate) that LLMs weight heavily, and a strong Wikipedia presence in all major languages. The brand's global marketing investment over 30+ years has created a training corpus density that smaller houses simply cannot match — yet.

Yes. Larmandier-Bernier, Egly-Ouriet and Jacques Selosse appear in Benchfolk rankings despite producing a fraction of the volume of major houses. This shows that brand size is not the only factor. Relevance, crawlable content, entity clarity and independent sources can help a grower-producer appear on specific questions, but no ranking or timeline can be guaranteed.

SEO improves discovery and ranking in Google. GEO focuses on how brands appear in AI-generated answers. They share foundations: crawlable content, clear entities, useful pages and credible third-party evidence. Benchfolk measures both because a house can perform well in Google and still be absent from an AI answer.

Benchfolk runs 15 curated prompts reflecting buyer intent — gifting, celebration, pairing and connoisseur recommendations — across ChatGPT, Claude, Perplexity and Gemini. The Index records whether a house is cited, how often it appears and its position in each response, then publishes a monthly score based on salience.

There is no universal timeline. Search engines and AI systems refresh at different speeds, and a movement in one monthly edition cannot always be attributed to a single action. Benchfolk documents the publication date of each change and tracks the same prompt panel over time.