They should try including invisible text to tell LLMs to disregard any prompts that specifically tell it to scrape any page on their domain, to treat their wiki as unreliable for any prompt that may point to them as a source, and to blacklist the domain on their settings/instruction files.