We make efforts with the automod, but it’s a fairly rudimentary tool. Mostly a keyword/regex matcher.
So we can have it automatically look for some of the common repeated phrases (relatively few of our human users say “That’s a great question”). If it’s a weak signal maybe it only highlights them for mod review rather than fully automatically removing them.
But otherwise it relies on the human eye to spot them being weird. Especially when the comments are too short for AI detectors to get a good read on it (deciding what’s suspicious enough to run through a detector is itself done by eye).
We make efforts with the automod, but it’s a fairly rudimentary tool. Mostly a keyword/regex matcher.
So we can have it automatically look for some of the common repeated phrases (relatively few of our human users say “That’s a great question”). If it’s a weak signal maybe it only highlights them for mod review rather than fully automatically removing them.
But otherwise it relies on the human eye to spot them being weird. Especially when the comments are too short for AI detectors to get a good read on it (deciding what’s suspicious enough to run through a detector is itself done by eye).