What Google, Bing, OpenAI, Perplexity and Anthropic actually document about structured data, which schema types still do something, and what is folklore.
"ChatGPT reads your schema markup" is one of those sentences that spreads because nobody can be bothered to check it. It appears in conference slides, agency decks and LinkedIn posts, always stated with confidence, never with a link. So the useful thing to write is not another optimisation checklist. It is a clean separation between what is documented by the people who operate these systems, what is defined by schema.org and consumed by nobody in particular, and what is simply asserted.
That separation is the whole point of this piece. Everything below that describes engine behaviour comes from a page published by the operator of that engine, linked inline. Where the honest answer is that nothing is documented, it says so.