The current upvote/downvote mechanism seems more than adequate to address these concerns.

If someone thinks an "I asked $AI, and it said" comment is bad, then they can downvote it.

As an aside, at times it may be insightful or curious to see what an AI actually says...