Sounds like an interesting topic for an HN discussion to me...
It's a mostly settled topic for now, you can find many moderation comments about it - don't use generated content in comments. The problem is that the overwhelming majority of LLM generated comments are not checked for accuracy, they're just random summaries of articles or unverified pastes of whatever the model spat out in response to a prompt. Even though you're trying to do this 'right' - checking and/or editing the output, disclosing the LLM use for the relevant part of the comment, etc - it's too difficult for anyone to tell the difference.
It's a mostly settled topic for now, you can find many moderation comments about it - don't use generated content in comments. The problem is that the overwhelming majority of LLM generated comments are not checked for accuracy, they're just random summaries of articles or unverified pastes of whatever the model spat out in response to a prompt. Even though you're trying to do this 'right' - checking and/or editing the output, disclosing the LLM use for the relevant part of the comment, etc - it's too difficult for anyone to tell the difference.