I can't really answer that because I haven't actually tried to use an LLM on any part of the process. The vast majority of the process is semantic markup using

I can't really answer that because I haven't actually tried to use an LLM on any part of the process. The vast majority of the process is semantic markup using (x)html and proofreading. The markup process could, I guess, use an LLM, but most of it is already automated using regex and linting.