Trained data vs web search (the gap)

AI engines answer in two distinct modes. In trained-data mode the model relies purely on what it internalized during training — knowledge with a cutoff date, weighted by how prominent your brand was in the training corpus. In web-search mode the engine retrieves live pages and synthesizes them, so your current site and recent coverage matter directly.

The gap between your scores in the two modes is diagnostic: a brand that scores well with web search but poorly on trained data is invisible whenever the engine doesn't search (many answers) — a long-term authority problem. The reverse pattern (strong trained, weak web) suggests your current site is hard for engines to retrieve or cite. BrandGEO audits both modes per engine and reports the gap.