There have been a number of flags raised recently regarding content that may have been generated by an LLM. It would be helpful to have a discussion about how these should be addressed in our community.

I find it difficult to determine whether content is LLM generated, and I prefer to assume good faith from users, but I also want to maintain the high quality of content in our community. Any interaction must adhere to our code of conduct and treat every user with kindness and respect.

Uncited use of LLMs is prohibited by our network-wide inauthentic usage policy. These tools should not be used in our community, but we still need to determine how to address potential use given that it is difficult to determine with high confidence when these tools are being used.

2 Answers 2


I'd say the source of a contribution normally shouldn't matter at all. If it's valid, relevant and correct: it's a good contribution - regardless of whether it was written by a top contributor, a newcomer, an LLM or a bunch of drunk monkeys throwing dice onto keyboards. The same (opposite) goes for bad content: it will get fixed or removed, no matter where it came from.

My proposal: so long as the amount of content awaiting moderation does not exceed the site's moderation capacity and the quality of submitted content does not fall noticeably, let us leave the status quo as it is. But if any of these happens, we should be ready to analyze the phenomenon and protect the community from its effects. For that case there are plenty (maybe even too many) online tools that purport to recognize texts generated by LLMs and larger sites must have more experience regarding this.

  • 1
    This makes sense to me. We'll plan for now to not take action regarding whether or not content is LLM generated as long as it meets our other quality standards.
    – jncraton Mod
    Commented Jun 25 at 11:39

I agree with Zovits, but wanted to add the team at SE had previously published their results around testing how good people and tools were at identifying AI content, and it wasn't good.

Lots of false positives, especially targeting people for whom English wasn't a first language. Their tests involved mix of older, pre-ChatGPT release content and newer content as a reasonable baseline.

  • 1
    Thanks for sharing those results. They provide very helpful context.
    – jncraton Mod
    Commented Jun 25 at 11:37

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .