The use of AI has been in our guidelines for ages, the issue was a growing number, of typically new, users using it to not just reply to others but to create overly structured and formal topics, that are more like a thesis than communication. It affects the tone of the site and engagement. It was becoming time consuming to moderate for a small team, and some people really didn’t like being moderated and then it becomes unpleasant and argumentative. I chose to create an AI agent in discourse (based on gpt 5.5) and use an automation to scan topic creation by new members in key categories, evaluating for signs of AI generation. This triggers an automatic flag and a polite explanation for why it was flagged, references to guidelines and some advice for better engagement. Mods can still release a false flag or borderline case. It has been very accurate and has driven the right behaviour while using way less time, and avoiding conflict.
We allow rate limited scraping from the major tech companies but have blacklisted the rest who were responsible for an enormous number of connections pushing us over usage limits. Like most bad actors on a site, it tends to be a small number of people who disproportionately consume the time and resources.