AI Allegations Thread

[MourningWarbler]

I'm bowing out of this convo (and this account!) soon. So carry on if you want.

 

[MediocreAuthor]

I think it would be very helpful if each person who posts here would include any grammar/writing aid programs they use, which could possibly help identify why everyone is getting so many false positives. Like the following: (feel free to copy/paste this as a template, if you like)

---

Number of rejections:

Grammar aids used:

Time spent pending:

Number of Moderator responses:

Length of Story:

Number of previous, unaffected stories:

Voice/Tense used:

---

If there is more relevant info, you can always suggest it, and I can add to template, if you deem that useful.

I use Pro-Writing Aid, but I've never had an AI rejection, but then again, I never use their automated suggestions for anything but punctuation and spelling, etc.

PWA suggests tone changes to improve clarity sometimes, but I think those may be tripping the AI detection software that Lit uses.
Also, I have a feeling that certain tenses and voices (like third person past) might register a false positive more than others; however, that is a complete unsubstantiated hypothesis.

Good luck with the problem, and hopefully it gets sorted out. I hate getting rejections... It's the worst.

 

[FeralSmile]

This might be helpful: https://contentwriters.com/blog/how-to-avoid-an-ai-detection-false-positive/

Generally speaking, AI detectors are likely to falsely flag content as being written by AI if it has simple sentence structure, uses common words, and is, in a word, predictable. LLMs function by selecting the most likely words and phrases to complete a given sentence, and the more your writing looks like their output, the more likely AI detection is to flag your output as being akin to what an AI would generate.

So break it up. Some sentences should be compound-complex, with sophisticated grammar and nuanced examinations of topics, such that an AI likely wouldn't choose them as likely or probable. And some should be short and... tight. I mean, AI would expect "sweet" there, right? If it's predictable, and if you write using cliches all the time, then yes, you're more likely to flag that AI.

 

[FeralSmile]

WE HAVE NO IDEA IF AN AI DETECTION BOT IS BEING USED. Some who have been accused believe that the mod is using her own judgment on this. And AI detection bots have been proven to be notoriously inaccurate. Also, why should we change our writing style, which in my case, was deemed just fine for 35 previous stories, in order to placate a bot? We writers affected by this are tired of being accused that this is happening because our writing isn't good enough.

Is your writing being flagged as AI? If so, then here are things you can change to prevent that. You can fret and gnash your teeth and rebel of course, but I'm not really sure what your point or goal there is.

Write more like a sophisticated human who is fluent in English, or don't, and face the possibility that your stuff will be rejected as AI, either by a human or by some script. I'm not picking a fight here, but trying to help. Do with that as you wish.

And, in your defense, and to be clear, it seems quite possible that Ernest Hemingway's writing would be rejected on Literotica, one way or another. Is that wrong? Is that ok? That's not really relevant, is it? I'm literally just telling you how you might better avoid this problem, and you're telling me you're too proud to do so. OK, I respect and understand that! But the outcome is some of your stories won't be published.

 

[FeralSmile]

Your "help" is absolutely useless. Take a look at my scores; every single one of my stories has a red H beside it. My writing is obviously good enough for the readers who enjoy my work here. How about you share your scores, since you're so quick to judge my writing without reading it?

Also, please note the thread title. This is for those specifically affected. Have you been? No? Then maybe keep quiet and take your advice elsewhere. Or catch up on the numerous other threads about this, where your kind of "advice" has been offered again and again and again.

I'm not judging your writing at all. As I specifically noted, Hemingway (usually accepted as a damned skilled writer) would fail these tests. So there's no need to get so defensive when no insult was made or intended.

I'm trying to help you avoid a problem with false AI flagging. And you're free to look up my story scores on your own if you wish. The link is in my signature.

 

[FeralSmile]

The problem with this is that these are stories that already have approved parts posted. If one part of the story is flagged for AI, it stands to reason that the rest should be too because it'll be bizarre for someone to start writing something and then switch to AI partway through when the AI wouldn't understand the characters the same way the author would. My second part got flagged for AI but reinstated after one resubmission. I'm sitting at 4 with my current part but the writing style hasn't changed between the parts.

There are a lot of inconsistencies with the way this is being handled and we can't guarantee that making adjustments to our writing would make the blindest bit of difference. We also shouldn't be forced to change the way we write because of something like this, especially because AI detectors aren't reliable and if Laurel is judging it herself, it becomes entirely subjective.

All we're looking for here is some sort of answer or explanation as to why this is happening and what the logic is behind these rejections because we haven't been told anything. The accusation of AI usage can hit anyone at any point, including you.

That's fair. I honestly wasn't trying to insult or demean anyone or their writing, but to offer input about how things get flagged as AI and how to avoid that.

And I hear you. Writing style isn't arbitrary. Writers have a voice. And if you're being flagged as AI it's not a trivial matter to change that voice. I understand why a writer would prefer not to change that in the face of AI accusations.

 

[FeralSmile]

I apologize for growing impatient; this is something I and others have been trying to resolve without success. But you did tell me to "write more like a sophisticated human who is fluent in English." That's a clear attack on my writing skills. And I see you have high scores here as well. Congratulations. Just know that absolutely will not protect your work from being targeted.

You're telling me to change my entire writing style, which readers have come to enjoy, in order to try to appease some kind of faulty AI detection system, be it human or bot. No, I'm not willing to do that. I write the way I write, and in being rejected after 35 stories of mine were deemed acceptable, I'm now being falsely accused of plagiarism. Instead of telling us to like it or lump it, as many others have, you can bow out of this conversation with the knowledge that all you're saying has been said many times before.

You're right, and I apologize. I phrased that poorly and I can see why you would take insult. What I meant to say was that varying your sentence style and structure, along with your word choice, can cause both human and script tools to recognize your voice and style as human and not AI. Writing with a specific voice, in some cases, will lead to flagging. And that's absolutely not a judgment of quality. Again, I sincerely believe that Hemingway would be flagged as AI for his short declarative sentences.

I'll bow out now. I was trying to help, and I apologize again for having offended. Best of luck to you in your writing here. I sincerely hope you solve this problem.

 

[nice90sguy]

I tried to find some successfully submitted stories by you -- have they ALL been rejected? Or is this your first attempt? I'm really curious as to whether there's something that would be cause a HUMAN to raise a false-positive on your writing. Some stylistic feature that stands out, for instance.

I've been posting stories here for more than 20 years, and I'm on first name terms with Laurel (her first name is "Laurel"), so I doubt if I'd ever get rejected for that reason, but I could easily imagine getting rejected for a first story here due to some resmeblence between my writing and AI-generated stuff.

 

[AwkwardMD]

I'll read someone's story rejected for AI. PM me.

Just one, though. I have limited spare time and focus.

 

[nice90sguy]

Can you link your stories? I couldn't find any using lit's shitty search tools

 

[AwkwardMD]

Thanks for this offer, but I have 35 stories published here which posted without any issue whatsoever. All have been highly rated by readers, so it's not a question of quality. I don't want to change my writing style in order to try to avoid whatever AI detection system is being used here.

35 stories in 6 months is approximately 1 story per 5 days. I looked at your story list, and they are 7k words, 9k words, 12k, 22k words.

Are you publishing a backlog of completed works, or is your output just that insane?

 

[AwkwardMD]

Oh, so you're going to start questioning writers who publish at a fast rate? Because there are plenty "established" and respected writers here who do so. Yeah, my output is that insane, because I'm a veritable recluse, and I also went through some heavy personal shit this year, so writing here was a great outlet. Didn't know that would make me suspect as well.

Pump the brakes. You are inferring tone that I did not imply.

 

[AwkwardMD]

Okay

 

[AwkwardlySet]

Oh, so you're going to start questioning writers who publish at a fast rate? Because there are plenty of "established" and respected writers here who do so. Yeah, my output is that insane, because I'm a veritable recluse, and I also went through some heavy personal shit this year, so writing here was a great outlet. Didn't know that would make me suspect as well.

When I first started writing, my output rates were about 1.5k -2k words per day or so, and I continued writing at that rate for about a month and a half before slowing down. I had a lot of time then and even more enthusiasm. Your output rates aren't that strange for someone who is very motivated and has plenty of time, but don't get unnecessarily paranoid when people express surprise at that speed, because it is quite remarkable.
We might be all writers here but believe me, when people want to accuse you of something, they won't be too subtle about it.

 

[onehitwanda]

No, I'm not going to pump the brakes. Since I've begun this thread, I've had my writing attacked, and now I'm being questioned for writing quickly.

Dude, you really need to take a breath. Nobody here has attacked you, they have asked questions which you have then interpreted as attacks. All you are going to do is alienate anyone who might be inclined to be helpful.

Can you link your stories? I couldn't find any using lit's shitty search tools

https://www.literotica.com/stories/memberpage.php?uid=7145027&page=submissions

 

[AwkwardlySet]

I just checked your profiles and yeah, many stories are gone. Laurel definitely suspects the use of AI, I would say. I suggest you PM her and make your case. I have no better advice to give you as this is not common occurrence on Lit.

 

[nice90sguy]

Laurel is perfectly entitled to refuse to publish a story that looks like it was written by an AI, whether or not it actually was. I know it must be galling to find one's personal style has been appropriated by Chat GPT , but I suggest you try to write something Chat GPT couldn't do. Consider it a challenge.

 

[AwkwardlySet]

I've just noticed that you mentioned copyright? That doesn't imply AI usage, that implies that your stories were reported for plagiarism, no?

 

[countdowntolov3]

I've just noticed that you mentioned copyright? That doesn't imply AI usage, that implies that your stories were reported for plagiarism, no?

A site owner like Laurel would see AI use as a copyright risk. Because the AIs are trained on huge numbers of copyrighted works, its currently unclear (from a legal perspective, as in, the court cases to set the precedent haven't happened yet) if publishing something from an AI is a copyright violation of the works used to train the AI. There might be artistic reasons to not want a lot of AI generated content. But the fear of a bunch of lawsuits is what has a lot content sites trying to avoid AI content right now.

 

[AwkwardlySet]

A site owner like Laurel would see AI use as a copyright risk. Because the AIs are trained on huge numbers of copyrighted works, its currently unclear (from a legal perspective, as in, the court cases to set the precedent haven't happened yet) if publishing something from an AI is a copyright violation of the works used to train the AI. There might be artistic reasons to not want a lot of AI generated content. But the fear of a bunch of lawsuits is what has a lot content sites trying to avoid AI content right now.

Ah, thanks. I forgot about all the authors who were suing OpenAI.

 

[Euphony]

when people want to accuse you of something, they won't be too subtle about it.

Self appointed Lit town criers' rage *intensifies*

 

[MillieDynamite]

Most AI checkers give you a precentage of probability that text is written by AI. I'm curious at what point Laurel, or whoever is checking for AI, determines it's too high a probability to allow it on Lit. It doesn't sound like we know what percentage is too much from what has been bandied back and forth. When one of the writers checked his work, he seemed to proudly proclaim that his work had a 70-something probability that it was AI-generated. Well, his stuff failed to be posted, so we can assume that 70% is a no-publish for Literotica.

 

[EmilyMiller]

This might be helpful: https://contentwriters.com/blog/how-to-avoid-an-ai-detection-false-positive/

Generally speaking, AI detectors are likely to falsely flag content as being written by AI if it has simple sentence structure, uses common words, and is, in a word, predictable. LLMs function by selecting the most likely words and phrases to complete a given sentence, and the more your writing looks like their output, the more likely AI detection is to flag your output as being akin to what an AI would generate.

So break it up. Some sentences should be compound-complex, with sophisticated grammar and nuanced examinations of topics, such that an AI likely wouldn't choose them as likely or probable. And some should be short and... tight. I mean, AI would expect "sweet" there, right? If it's predictable, and if you write using cliches all the time, then yes, you're more likely to flag that AI.

So my sesquipedalian tendencies, science / math jargon, and fondness for both ultra short and lengthy parenthetical sentences is my super-power!

Cool! I knew they would come in handy one day .

Em

 

[MillieDynamite]

I've always said that AI detection isn't any good. Even a human can write the way AI does as a style. If that style is good or bad, I'm not qualified to say. But AI has wild changes that make no sense. Introduces characters with a name, never mentions them after a couple of paragraphs, and gives you no clue as to where they went. (Think only AI does this, no Stephan King does this as well. Or worse gets you interested in someone and never mentions them again.) Writers do that as well, and why they give characters a name if his only giving you a drink at the bar, I haven't a clue. I just don't see any real way that AI can be detected by humans or software to any degree of certainty.

Because you've been very helpful, just want to let you know that I posted my stats from another one of my accounts. Same writing style, same fast writing pace. Take a look before the account's deleted, because I requested it be deleted. Last published there in late 2021. No ChatGPT available then. So that would definitely indicate that my account is being erroneously flagged now. Whatever system is being used is FLAWED and INACCURATE.

 

[SamBomb]

Yeah AI can do somethings well, but they suck at writing stories.