I'm hoping the in the future there will be an LLM that can somehow map text back to its source. Current LLMs can kinda pull it off if you ask them, but afaik they aren't trained specially for this task.