This is only if the perlmonks text was swept up in the training data. It's probably in our best interest to make perlmonks more downloadable so that this body of information is available to LLM tools. People might actually decide to use or not use perl for a task based on how well ChatGPT can answer questions about it. Lately I've been asking it a bunch of questions about Vue3 and amazes how useful the answers are (as a search engine, it still doesn't write accurate code).