Less than half of the spam for sure… Yet. I’ve been getting 5-10 spam accounts a day for years.
It adds a bit of noise.
The AI spam always looks like this: a well-written, not too formal, not casual post that properly makes use of the context of the forum’s interests, or topic in which it’s posted. The post clearly looks legit, but they tend to contain a random spam link in one of the post words, which is immediately noticed by the community, and flagged by users.
Users report and flag. Also, these post often fall under the Min first post typing time treshold, and are automatically detected as spam by Discourse.
I’d be very interested in a data explorer query that would return the estimated AI spam posts % over a certain duration. Obviously, the criteria for an AI post are very difficult to narrow for a SQL query, but in my case I think I could use keywords. Generic spam usually doesn’t use keywords specific to my community’s interests, while AI posts do use them.
Since AI posts are currently always reported by users, allowing them to be returned as reviewed (rejected) items in a SQL query.