We took 40 product pages from consenting brands, asked five answer engines the same buying question for each, and logged every citation. 23 pages were never cited by any engine. Eight were cited by three or more.
The eight winners were not the biggest brands. They had a spec table the engine could lift verbatim. They answered the exact buying question in the first paragraph, in plain language. And their prices matched what the engine found everywhere else.
Engines cross-check. A page that disagrees with the rest of the web about its own product loses to a reseller that agrees.
The full run, prompts, and per-engine scoring are in the dataset. The experiment repeats quarterly. The next run adds agent checkout flows.