No AI Content Rule

[surferfb]

This rule is, as far as I can tell, never enforced (unless someone accidentally pastes in the prompt too):

Don't post text produced by artificial intelligence ("generative AI"). Even if you use generative AI to research an issue, posts must be in your own words and represent your comments or opinions. Exception: You can post AI-generated text if you are giving limited examples, clearly identified, in a discussion thread about the topic of generative AI.

I understand this isn't always cut and dry, sometimes it can be difficult to determine if something was AI-generated or not, I personally think too many things get called out as being AI generated when they're not, and I generally would want the moderators to give the user the benefit of the doubt. But whenever I report posts that are clearly, unambiguously, written entirely by someone using an AI tool I get told "no moderation action is warranted." If the thought is "it's too hard to know for sure", then maybe we should scrap the rule?

I am talking about things like one post:

• following the three paragraph thesis-rebuttal-clincher structure,
• rhetorical contrast setups ("this is not X, but Y"),
• emphatic mid-paragraph exclamation points (!),
• vague consensus from "experts" ("That's why [unnamed professional class] keep calling it [damning label]",
• tidy conditional-disproof logic ("if x meant y, it wouldn't do z when..."),
• a closing one-sentence verdict that wraps everything up in a tidy little bow.

Since I assume I cannot share examples publicly (happy to DM the mods if that is helpful), here is a "limited example, clearly identified, AI-generated" non-tech example following all the hallmarks of a post I recently reported and was told wasn't against the rules:

Unlimited PTO isn't a generous employee benefit. It's a burden enforced by total control over workplace norms and social pressure. Workers never chose to self-police their own time off in a fair system; they were locked into it because the policy eliminated accrual tracking and removed any real accountability from management! That's why labor researchers keep calling the arrangement a cost-shifting scheme.

If the policy reflected genuine employee wellbeing, it wouldn't result in workers taking fewer vacation days than they did before. Calling it a trap is a fair description of a benefit designed through employer control rather than actual generosity.

As someone who works with AI all day long, it's as clear as day that something like that was 100% generated by AI. But whenever I report these things, I get told "nope, doesn't break forum rules."

Again, I get that it's a hard problem, and I don't claim to have an easy answer. One thought I had is "maybe user gets a warning" a few times where the Mods say "we're not going to moderate the post, but this has many aspects generally associated with AI-generated content, please remember to write things in your own words" and if a user gets them repeatedly then moderation happens? Not trying to create additional work for mods, but right now it feels like I shouldn't bother reporting these.

 

[S.B.G]

I have advocated for a change to the rule whereby AI content is allowed as long as it is in quoted form within the post and clearly identified as AI content.

Example post:

Hey, I was trying to find the differences between the M3 and M4 MacBook Air's and this is what AI gave me:

Per MacRumors, the M4 MacBook Air is a focused but incremental generational step. The chip moves from an 8-core CPU to a 10-core CPU (four performance and six efficiency cores), with clock speed increasing from 4.05 GHz to 4.3 GHz and transistor count rising from 25 billion to 28 billion. Memory bandwidth jumps from 100 GB/s to 120 GB/s thanks to a shift to LPDDR5X memory, and the base RAM configuration starts at 16GB rather than 8GB, with a new 32GB option at the top end. The Neural Engine doubles its throughput — from 18 to 38 trillion operations per second — which has direct implications for Apple Intelligence performance. A significant practical improvement is the external display situation: the M3 could only drive two displays in clamshell mode, while the M4 supports two external displays with the lid open. The webcam also gets a major upgrade, going from a standard 1080p FaceTime HD camera to a 12MP camera with Center Stage and Desk View support. The design and display are unchanged, Space Gray is replaced by Sky Blue, and the starting price drops back to $999. MacRumors' overall take is that neither generation represents a dramatic leap — the primary audience is people upgrading from M1-era machines or older, and moving from M3 to M4 specifically is hard to justify unless dual-display support or the improved camera are meaningful to your workflow.

In this scenario there is original user content in the post, and AI content using the quote function and I spelled out that what follows is AI content.

I think that is a perfectly acceptable use of AI on the forums, and gives the users latitude to use AI without running afoul of the rules.

 

[poorcody]

I see nothing wrong with your AI example being posted.

Let's make it clear the purpose of banning AI. If it's to avoid mass bot-like spam, okay. If it is just for its own sake, I don't think it should be banned. A lot of AI produced text is quite informative. Especially compared to what a lot of humans post.

 

[Weaselboy]

Since I assume I cannot share examples publicly...

Correct... please don't out the other member.

If you are referring to the Apple fees post you reported this morning, that post did not jump out at me (like some AI posts do) as AI. I also looked at that member's post history and did not see anything that looked like a history of AI comments.

I also ran the post through two different AI detectors, and both said 0% chance the text was AI generated (I recognize these detectors may not be perfect).

We have moderated quite a few AI posts since that rule was enacted, so please don't be discouraged by one we rejected and continue reporting suspected AI posts.

 

[Nermal]

Let's make it clear the purpose of banning AI. If it's to avoid mass bot-like spam, okay. If it is just for its own sake, I don't think it should be banned. A lot of AI produced text is quite informative. Especially compared to what a lot of humans post.

My personal opinion is that people come to a discussion forum to have discussions with other people. If they wanted to talk to an AI then they'd talk to an AI.

With my moderator hat on, like Weaselboy says above, we do regularly moderate AI posts. Sometimes they can be subjective, and when there's no consensus between moderators then we let the post stand.

 

[surferfb]

Correct... please don't out the other member.

If you are referring to the Apple fees post you reported this morning, that post did not jump out at me (like some AI posts do) as AI. I also looked at that member's post history and did not see anything that looked like a history of AI comments.

I also ran the post through two different AI detectors, and both said 0% chance the text was AI generated (I recognize these detectors may not be perfect).

We have moderated quite a few AI posts since that rule was enacted, so please don't be discouraged by one we rejected and continue reporting suspected AI posts.

Thanks for the feedback! Yes, this was prompted by the post I reported this morning, but I've never had a report on AI use accepted, so it's not just like it was a one time thing. Obviously it's a judgement call and if you don't think the post in question reads as AI-written then you definitely shouldn't be moderating it.

That said, I'm honestly shocked your tools didn't think it was AI generated. I am pretty sure I could name not only which AI tool, but also the model used - and the normal tool I use for checking flagged it as 100% AI generated. If you have time, I'd be curious to know if the example I included in the first post in this thread gets flagged as AI-generated (because it absolutely was).

But as I said, I think if it's questionable you should give the user the benefit of the doubt. I've actually had something I written falsely accused of being AI generated (not here), so I get it, and I think, in general, the less moderation the better. Thanks again for the quick response! I know you mods have a hard job and don't want to be adding extra work. 🙂

 

[Weaselboy]

and the normal tool I use for checking flagged it as 100% AI generated.

If you don't mind, would you please PM me the link to the tool you use. Always willing to look at new tests for this.

 

[rb2112]

• rhetorical contrast setups ("this is not X, but Y"),
• tidy conditional-disproof logic ("if x meant y, it wouldn't do z when..."),
• a closing one-sentence verdict that wraps everything up in a tidy little bow

slightly off topic, but I have recently unfollowed many of my content creators that write like the examples you listed, ESPECIALLY these quoted. I don't know if they are AI. But I think they are an overused rhetorical technique....when everyone is doing it, gets stale fast.

 

[maflynn]

My concern is that I frequently use AI to rewrite my emails/messages/texts so I don't sound like pre-schooler.
AI is also useful to collate disparate information, while it needs to be cross checked, its a useful tool, I hate to see something blocked completely.

 

[poorcody]

My personal opinion is that people come to a discussion forum to have discussions with other people. If they wanted to talk to an AI then they'd talk to an AI.

I don't feel that way... I come for the best information (when not joking around or otherwise having fun).

AI output should be attributed just like posting information from other people or sites, but I definitely wouldn't want it banned when it makes valuable contributions. And frankly if it just cleans up people's writing -- I think that's good too. The higher the quality of what we ingest, the better we become as thinkers and writers.

 

[Black_Mage]

As long as I am allowed to use Grammarly. I sometimes type fast to get my ideas recorded and clean up the mess later with the help of Grammarly.

 

[Black_Mage]

My concern is that I frequently use AI to rewrite my emails/messages/texts so I don't sound like pre-schooler.
AI is also useful to collate disparate information, while it needs to be cross checked, its a useful tool, I hate to see something blocked completely.

The only problem I have ever seen with your posts is that you confuse when to use then and than. Use then when you are talking about the order of events, and use than when you are comparing things. I used to use Windows then I switched to Mac. I'd rather use MacOS than Windows. Grammarly can help with that, and it's free to use.

 

[maflynn]

The only problem I have ever seen with your posts is that you confuse when to use then and than.

My wife gets on my case for that as well. I think I'm missing that then/than gene, because I simply cannot discern the use. My concern is that I'm communicating with executives, and senior directors on topics, so I tend to pump my correspondences though AI.

 

[GGJstudios]

My concern is that I frequently use AI to rewrite my emails/messages/texts so I don't sound like pre-schooler.

Trying to circumvent the minimum age limit on this site? Your post count DOES make more sense if you factor in the energy level of an average pre-schooler!

My wife gets on my case ...

OK, maybe not, then! 😉

 

[theluggage]

I think that is a perfectly acceptable use of AI on the forums, and gives the users latitude to use AI without running afoul of the rules.

In a sense, it's worse. "per MacRumors" implies that the data is sourced from MR, but it's unclear whether the passage is a quote from MacRumors or the AI's 'interpretation'. What specific article did the information come from from? URL? Maybe it just put "Per MacRumors" in there because the model threw up a statistical connection between summaries of Apple products and citations to MacRumors.... because that's what LLMs do.

AI simply can't be trusted to get all of the facts correct, and it obfuscates the original sources - by the time you've gone through and checked all the references exist, are correctly quoted and haven't been hallucinated, you might as well have done it the hard way.

Alternatively, you could have just posted one URL:

M2 vs. M3 vs. M4 MacBook Air Buyer's Guide: 25+ Differences Compared

Apple just announced a new MacBook Air, introducing its latest chip, an enhanced camera, and a striking new color option. Simultaneously, it discontinued the two previous generations. So how does the new model compare? Before the introduction of the latest model, the 13-inch M3 MacBook Air...

www.macrumors.com

...that takes 10 seconds to find with regular non-AI search (or not long to find manually in the MR buyers guide) and lets people read directly from the source, follow onward links, verify the source and decide whether they trust it, even read the comments to see what criticism it has received.
(Edited: corrected one typo)

 

[Grey Area]

A lot of AI produced text is quite informative.

I disagree on this. AI-generated text has no value, it is basically disconnected from information. AI has no understanding whatsoever of the world, nor of the words it strings together. It cannot distinguish between fact or fiction. Moreover, it is not even comparable to some human who only knows the world from books: AI has no understanding of contradictions within its "knowledge", it does not even understand what true or false mean. It also has no meta-cognition. Notably, it has no human-like means of concluding that "I don't know" is the appropriate response, as it cannot "think" about what it "knows". In that sense it is less intelligent than a traditional database.

People sometimes complain that AI "hallucinates", but that is all it can do, all its texts are hallucinations, some just happen to correspond to reality. It is a mindless automaton geared to generate word chains that superficially resemble human text. Reading AI-text is a waste of time. One would have to fact-check every single word before one could treat it as information. And there is always the risk of something slipping through, because it is so easy to anthropomorphize these machines.

 

[adrianlondon]

Agree with the recent posts.

I used to look at Google’s AI text at the top of searches, think it was useful, and click on the links it said it got the info from.

(I did this as I believe the actual sites containing the info deserve the hits.)

The linked sites usually had info that contradicted the AI invented summary.

I’ve stopped looking at it.

 

[rb2112]

I disagree on this. AI-generated text has no value, it is basically disconnected from information. AI has no understanding whatsoever of the world, nor of the words it strings together. It cannot distinguish between fact or fiction. Moreover, it is not even comparable to some human who only knows the world from books: AI has no understanding of contradictions within its "knowledge", it does not even understand what true or false mean. It also has no meta-cognition. Notably, it has no human-like means of concluding that "I don't know" is the appropriate response, as it cannot "think" about what it "knows". In that sense it is less intelligent than a traditional database.

People sometimes complain that AI "hallucinates", but that is all it can do, all its texts are hallucinations, some just happen to correspond to reality. It is a mindless automaton geared to generate word chains that superficially resemble human text. Reading AI-text is a waste of time. One would have to fact-check every single word before one could treat it as information. And there is always the risk of something slipping through, because it is so easy to anthropomorphize these machines.

Agree with most except to add I think AI generated text can have value, and it really depends on the situation and how much you are overseeing the generation process. Someone said it was like having an overeager grad student that you would have to check all their work. I have used LLMs to help me look at things in new ways, to find a reference that I have missed up until now, for a bit of value per work session.

But yes, I can glance at (too many) web pages these days that were a response to my query and see (Oh, AI generated this) and move on because the work I would have to do to eke out the info I need from the slop page is just too much. I think the internet is taking a temporary reduction from it's usefullness to simply queries and hope that (a) AI gen'd pages can get masked out or (b) AI improves or something else. Sadness.