Every prompt already has winners. If you are not named, someone else is. Competitive share is mention and citation against the brands the model already puts on that panel — not a mood board of rivals from last year’s SEO deck.
Suggested competitors are a starting list, not a verdict. The engine co-cites. That co-citation is the map. Pitching a publisher the model never uses is a different sport from earning the next URL on a domain it already likes.
Brand perception is the attributes models attach: fast, trusted, expensive, “for enterprises,” “for beginners.” Perception is not a truth claim. Content aims at what makes them recommend, not at a slogan you wish they would repeat.
A category benchmark is how the panel looked that month. It is not a trademarked index. Methods across tools are incomparable. Anyone selling a universal “AI Brand Score” as science is selling a label.
Position in the reply still matters. First named is not the same as buried after a warning. Sentiment still matters. Named and warned-off is not a win.
The useful meeting question is not “what is our rank in ChatGPT.” It is “who occupies this answer, with which URL, in this market, on this date — and what do we publish or pitch before the next slice.”