Is casting R/A flags on blatant, but uncited, LLM-generated content acceptable?

Question

This is about LLM-generated content that wouldn't otherwise be R/A, not abusive content that happens to be LLM-generated. Also, spam is still spam, even if it's AI-generated.

For sake of discussion, I'll define "blatant" AI-generated content as "clear to a reasonable user with no access to the heuristics". While I cannot disclose the heuristics, an answer starting with "As an AI language model ..." would qualify as blatant here.

Per What are the “spam” and “rude or abusive” (offensive) flags, and how do they work?

A post should be marked as rude or abusive [...] if it contains [...] abuse of the community or system, i.e., a clear violation of the Code of Conduct.

I consider LLM-generated content an abuse of the system and uncited content is a CoC violation per Inauthentic Usage.

As a rule of thumb, everything that would be out of place in polite discourse is rude or abusive.

If I asked someone a question, and they started their reply with "As an AI language model", I'd consider that impolite. If I messaged a colleague, and they posted a blatantly ChatGPT dump as a reply, I'd consider that rude, and absolutely out of place.

(under When not to use the flag) The post blatantly violates other site rules. This is what downvotes, close votes, and other flag types are for.

While it's debatable what "other site rules" means, I'd argue it's a CoC violation. We still red flag inappropriate content even if it's also off-topic. (Or, we shouldn't close R/A or spam posts, anyways)

Interestingly, there's also an MSO discussion about normal users (without diamonds) casting delete votes on suspected LLM answers. The conclusion there seemed to be "Normal users aren't bound by heuristics, so we can cast delete votes if we want, and disputes can go to Meta. It might be messy, but that's SE's fault here". It was given status-review, but I don't see a staff answer there, so IDK to what extent that's policy.

---

Specifically, I'm not asking if it's preferred or ideal - it's clear that [custom] mod flags are a better option. But, if I want something blatantly fake gone, and I don't want to wait, may I cast an R/A red flag under the rules for when to use red flags?

Answer 1

I think you answered your own question in quoting:

Do not use this flag if ...

• The post blatantly violates other site rules. This is what downvotes, close votes, and other flag types are for.

If LLM content is not allowed under a particular site's rules, then it should be flagged for Moderator attention.

There's good reason for this -- We want Moderators to see and be able to take AI-appropriate action on AI answers. This means that, when an answer from a user is determined by a Mod to be AI, they can:

1. Not only remove the answer, but also ...
2. (Sometimes more importantly) Provide a warning to the user that it is not permitted.
3. If it's a repeat offense, take appropriate additional action, such as potential suspension of the user.

R/A flags can prevent that from happening by allowing deletion of the post without a Moderator ever reviewing it.