Reference examples are sparse and chatgpt hasnt got much info

Chicken, meet egg. If a web crawler can’t find plenty of reference examples, then LLMs trained on web crawled data aren’t going to be very useful.