One more realistic option could be to have an "LLM browsing proxy" where you chat with an LLM via text, and it does the browsing and parsing and extracting, with links etc.