The "which sources the AI systems trust and cite" part is the one I'd actually pay for, but it only helps if you can split it by how the answer got produced. We run a job board in iGaming and ran 25 queries against Gemini earlier this month, and we showed up in exactly zero of them. That turned out to be two different problems wearing the same costume. Some assistants were answering from memory and simply had no record we existed; others were fetching live and choking on a JS-rendered page. The second one we fixed in a week with static HTML, markdown mirrors and an open API with no key, and the first is still a slow grind through other people's pages. When CitePrism shows a brand as missing, can it tell you which of those two you're looking at?
CitePrism always runs with search on, so there's no memory-only comparison today, but it does show per-answer citations, search queries, and gap analysis against competitors.