AI Rejection

[RedFireBrand]

But that’s not an invitation. Just saying that non sequiturs are an AI trait.

Emily

if thats an action someone takes, voluntary or not (so long as its not forced), its a non verbal invitation

 

[EmilyMiller]

if thats an action someone takes, voluntary or not (so long as its not forced), its a non verbal invitation

OK - not going to argue with you…

Trying to help.

Emily

 

[RedFireBrand]

OK - not going to argue with you…

Trying to help.

Emily

no worries. its fun to discuss things sometimes

 

[LadyLibido]

Personally, it bothers me to see how many users are getting smacked with accusations of using AI over the last couple months. (it's very clear to see from the Similar Threads suggestions here, right now) It's been inconsistent as to why this is happening, because some of the users that have been flagged don't even use services like Grammerly or CoPilot. Testing the samples provided in most publicly available AI detection tools does not flag them as AI, so it doesn't appear to be a matter of a detection tool causing a false flag, which means it was someone's personal discretion. And most frustrating of all, the moderation team isn't providing any information about what or why exactly something reads as AI. And the way the whole system is structured doesn't even afford the opportunity to talk through the process with a mod, the only way to do that is to directly contact Laurel or Manu which I can't imagine is desirable for either of them.

I saw Emily mention a suggestion specifically for AI detection flag disputes, and if this keeps up, that might be the only way to go.

That aside, this seems like a precedent even those of us that have never struggled with, should be concerned about.

 

[travelenman]

My story was also rejected based on the AI issue. I do use Grammarly for spelling and grammar, but Grammarly is not great with erotica sentence structures. This is not a complaint, and I accept the lit rules as I write these stories for the audience and with an average score of 4, I could hardly be considered good at this.

 

[TheRedChamber]

"Amy, what is going on? I can tell you aren't telling me everything?" Mel finally blurted out.

"It looks like you need more tanning oil," I said, attempting to change the subject.

I walked over to Mel's lounge chair, seductively swaying my hips, attempting to pull her attention to my body.

"But Amy," Mel responded.

It was clear she wanted answers, but I had none to give. I splashed oil onto my hands, rubbed them together, and brought them to Melonie's tight brown ass, hoping it would distract her.

Mel's bare ass felt divine underhand. I massaged the oil into her skin, making sure not to miss anything.

"You know this doesn't change anything," Mel said.

"Just shut up and enjoy it," I giggled.

I ran my slick hands up and down her body. Mel's tits flowed out of her tiny top, revealing a ton of cleavage as she lay on her stomach. I trailed the edge of her bikini, applying oil to her exposed breasts. Moans escaped Mel's lips as I teased her sensitive skin.

Honestly, I don't know anything about how AI works or how AI judges AI is other writing, so I'm just spitballing here and none of this may be worth a damn. I'm also looking at a tiny fragment which might not be representative of the whole.

The bits I've bolded I think are fairly generic chunks - strings of four plus words that are very commonly found together regardless of the genre of fiction and which might be likely candidates for an AI writing a script - once it choses the first couple of words, it's fairly likely that your next word is fairly high up the list of possible candidates.

The other thing I noticed was you use the 'comma -ing' clauses structure a lot - and I mean a lot. It's in nearly all of your non-dialogue sentences here. Again, no idea if it's related to AI-detection or not, but it might be worth being aware of stylistically anyway.

 

[PenumbralVision]

Hi everyone

I'm new here and my first story got rejected because of the same reason. Just wanted to thank you for your suggestions on how not to write like an A.I. Unfortunately, it seems like my first, admittedly clumsy, attempt at writing a short story for Literotica ticks all the boxes.

I have (and always had) a tendency towards using long sentences, a flowery vocabulary, formal language and an excessive amounts of fancy adverbs. Also, I prefer to describe situations instead of writing dialogue, mainly because my previous attempts at doing so ended up sounding weirdly unnatural.

English is not my native language, and I fear that might be part of the problem, since, in my experience, people tend to talk and write much more formal and elaborate when using a second language. Perhaps that's just me though.

 

[lovecraft68]

Just have shoddy grammar like me and it won't be a problem.

Use slang and things like "where ya goin'" and you'll have no problems.

AI writes too perfectly.

That's write, us hacks have had it right all along.

 

[PenumbralVision]

In a surprising turn of events, my second attempt at writting a story was published after barely three days, despite being very similar to the rejected one.

 

[PenumbralVision]

Yeah, it doesn't make any sense. One of my parts went through without issue but it was written the same as the other parts that were rejected

I wonder what software is being used to decide if a text is A.I written or not. It seems to decide pretty unpredictably.

Anyway, I guess I'll rework my first story and try submitting it again

 

[MillieDynamite]

I don't believe it's using Grammarly or any other aid. It's strictly the choice of certain words or phrases. What those words or phrases are, I couldn't say. I've been lucky, or I think I have, in that I've not had any issue of having anything sent back for AI.

Personally, it bothers me to see how many users are getting smacked with accusations of using AI over the last couple months. (it's very clear to see from the Similar Threads suggestions here, right now) It's been inconsistent as to why this is happening, because some of the users that have been flagged don't even use services like Grammerly or CoPilot. Testing the samples provided in most publicly available AI detection tools does not flag them as AI, so it doesn't appear to be a matter of a detection tool causing a false flag, which means it was someone's personal discretion. And most frustrating of all, the moderation team isn't providing any information about what or why exactly something reads as AI. And the way the whole system is structured doesn't even afford the opportunity to talk through the process with a mod, the only way to do that is to directly contact Laurel or Manu which I can't imagine is desirable for either of them.

I saw Emily mention a suggestion specifically for AI detection flag disputes, and if this keeps up, that might be the only way to go.

That aside, this seems like a precedent even those of us that have never struggled with, should be concerned about.

AI doesn't write perfectly; it writes poorly, some of which is perfect, some of which isn't. AI mines by scrubbing through written material. It's only as good as that which it copies. But it doesn't really do dialog well at all. The examples I've seen are stiff and sound boring and unrealistic.

Just have shoddy grammar like me and it won't be a problem.

Use slang and things like "where ya goin'" and you'll have no problems.

AI writes too perfectly.

That's write, us hacks have had it right all along.

 

[loerics]

I submitted a story like I have 68 other times with the same process. I write, toss it into Grammarly to edit for commas, remove tautologies, and other minor corrections, then i give it to my editor. He gives me his list of changes and I make the corrections. Then I submit. 68 times this hasnt been an issue. Now, suddenly, my story keeps getting rejected for AI 'rewriting' my work.

I manually disabled all AI features with Grammarly, just in case I was using something without knowing it. I went back to my rough draft and re-edited it, taking less suggestions from Grammarly, and redid my editor changes. Still got rejected.

I reached out to Laurel with admin notes and a direct message here and got nothing.

I am not using AI to write any amount of my story. I'm using spellcheck and an editor. Whatever method of detection being used is flawed. Getting a bland notice that tells me nothing of why my work got rejected does not help me. I understand the reason why a hard line needs to be taken with AI but after 8 years of uploads I think its safe to say I am not some hack with a bot.

I just had the same AI rejection problem and got a notice that is useless. I use Grammarly because I’m terrible with commas and plain lazy. Grammarly often recommends dropping extraneous words like ‘really’ just as a human editor would do. The most dramatic feature I use is rearranging phrases in a sentence for clarity. Oh and I also use thesaurus software. Every word in my 110 stories are mine.
I’m going to resubmit without changes and attach a strongly worded message.
My only alternative is to publish on ASSTR.
PS Let us know what happens with your rejected story.

 

[loerics]

Do you let Grammarly rewrite your sentences? If so, then that could be the problem.

How would they know? Grammarly doesn’t watermark the story. Word Properties doesn’t indicate that Gammarly was used.
I use Grammarly for spelling, commas, deleting extraneous words like ‘really’, and sometimes rearranging phrases in a sentence for clarity. No way did an AI write any of my stories. I experimented with an AI and the story was drival.

 

[Britva415]

How would they know?

I'm not sure it matters. I think what matters is that that is what the policy is, and you're clearly arguing for breaking the policy because you don't believe it's possible to get caught.

Ironic, since you are already getting caught. And blabbing in public about it is likely to get you more scrutiny from the site's publishing approver in the future.

 

[Britva415]

most of those who do (without using it to change sentence structure) are being rejected for AI

I don't think any of us have any idea how true that is, depending on who you mean by "us."

Do you mean "us" in this thread?
"Us" as in everyone getting rejected for AI, whether they're posting about it or not?
Or do you mean "us" as in all the people who are submitting stories every day?

 

[STrent]

I’m having a bizarre week.

Five rejections for AI, then it’s looking like five publications in a week including two of my repeat offenders and three new chapters.

Its puzzling!!!

 

[Elayne_Hawke]

So I got a question, (might be stupid) how do you know that something was rejected by an AI?

So when I first submitted my story, I did it in a group. The first five chapters bundled at once. Found out afterwards that was a mistake. Not everyone likes to read stories like that. So I decided to submit the chapters individually and my first chapter keeps getting sent back. First time, I left a note for the mods that I was separating the chapter, when it was rejected I saw a thing about changing title. Ok, did that, resubmitted and I got the same bloody response after it was rejected again. Its not a new story, as I said, I'm just separating the chapters. I'm about to say screw it and leave it the way it is.

 

[Helgamite]

So I got a question, (might be stupid) how do you know that something was rejected by an AI?

So when I first submitted my story, I did it in a group. The first five chapters bundled at once. Found out afterwards that was a mistake. Not everyone likes to read stories like that. So I decided to submit the chapters individually and my first chapter keeps getting sent back. First time, I left a note for the mods that I was separating the chapter, when it was rejected I saw a thing about changing title. Ok, did that, resubmitted and I got the same bloody response after it was rejected again. Its not a new story, as I said, I'm just separating the chapters. I'm about to say screw it and leave it the way it is.

I enjoyed them. having them all like that made it great to keep the story together and with chapters 6 and on being separate, not a bad idea as it can keep them coming back and makes the burden of writing a little more tolerable in the long run.

 

[ElectricBlue]

So I got a question, (might be stupid) how do you know that something was rejected by an AI?

We don't. That's the fundamental concern right now. People are getting rejection notices for their stories being suspected as AI written or AI assisted, but the site has not yet disclosed exactly how it arrives at that suspicion.

It appears to be a combination of bot detection and human eyes, but the site has not disclosed the process.

 

[loerics]

I'm not sure it matters. I think what matters is that that is what the policy is, and you're clearly arguing for breaking the policy because you don't believe it's possible to get caught.

Ironic, since you are already getting caught. And blabbing in public about it is likely to get you more scrutiny from the site's publishing approver in the future.

I have worked in Computer Vision AI since the 1970s. Currently, AI-generated images can be detected thanks to anomalies, such as hands with six fingers. Those errors will soon disappear. In my opinion, detecting AI-generated writing is a harder problem. Writing styles vary significantly from person to person. Current detection programs
Some

I'm not sure it matters. I think what matters is that that is what the policy is, and you're clearly arguing for breaking the policy because you don't believe it's possible to get caught.

Ironic, since you are already getting caught. And blabbing in public about it is likely to get you more scrutiny from the site's publishing approver in the future.

I hope you didn’t mean for your reply to be so condescending and sanctimonious.

Of course, it matters if there is a reliable way to detect AI content.

From my viewpoint, the only difference between the hardliners on this thread is that you haven’t had a story rejected. If Laurel is using AI detection software, it’s only a matter of time before you are flagged.

I have worked in Computer Vision AI since the 1970s. Currently, AI-generated images can be detected thanks to anomalies, such as hands with six fingers. Those errors will soon disappear. However, you’ll never get me to let a Tesla drive for me.

In my opinion, detecting AI-generated writing is a harder problem. Writing styles vary significantly from person to person. My comment ‘How would they know?’ is based upon how bad AI detection software is. Perhaps, Laurel doesn't use software and relies upon her considerable review experience. Sorry but that is just as flawed a process.

Some hardliners posting on this thread, suggest you need to become a better writer, or you shouldn’t use spell or grammar checkers. Well even though this is an amateur site, I got attacked by Grammar Nazi for my numerous comma errors. Now I use Grammarly for spell checking, grammar, and occasionally I accept recommendations to rearrange my phrases in a sentence. All the words I post are my own. I have never accepted Grammarly’s offer to improve a paragraph. There is no difference from the input I got from editors and beta readers.

I think you’ll understand when you get a rejection. A lot of authors with over a hundred posts have been rejected and don’t know why. I have 110 stores posted. If you say any of them have AI content, then you don’t know what AI is.

A major problem with AI systems is that they cannot explain how they make decisions. The best you can do is test it on additional data and look for patterns. You cannot “program” an AI system to look for things like 'perplexity' and 'burstiness.' All you can do is train it on a lot of data for which you somehow know the truth and look for patterns elsewhere. I find it amusing that people on this thread have added spaces to text and achieved better scores from their favorite AI detector.

Basically, I agree with the posters who called this a witch hunt. Be afraid. Be very afraid. You may be next.

 

[Helgamite]

I have worked in Computer Vision AI since the 1970s. Currently, AI-generated images can be detected thanks to anomalies, such as hands with six fingers. Those errors will soon disappear. In my opinion, detecting AI-generated writing is a harder problem. Writing styles vary significantly from person to person. Current detection programs
Some

I hope you didn’t mean for your reply to be so condescending and sanctimonious.

Of course, it matters if there is a reliable way to detect AI content.

From my viewpoint, the only difference between the hardliners on this thread is that you haven’t had a story rejected. If Laurel is using AI detection software, it’s only a matter of time before you are flagged.

I have worked in Computer Vision AI since the 1970s. Currently, AI-generated images can be detected thanks to anomalies, such as hands with six fingers. Those errors will soon disappear. However, you’ll never get me to let a Tesla drive for me.

In my opinion, detecting AI-generated writing is a harder problem. Writing styles vary significantly from person to person. My comment ‘How would they know?’ is based upon how bad AI detection software is. Perhaps, Laurel doesn't use software and relies upon her considerable review experience. Sorry but that is just as flawed a process.

Some hardliners posting on this thread, suggest you need to become a better writer, or you shouldn’t use spell or grammar checkers. Well even though this is an amateur site, I got attacked by Grammar Nazi for my numerous comma errors. Now I use Grammarly for spell checking, grammar, and occasionally I accept recommendations to rearrange my phrases in a sentence. All the words I post are my own. I have never accepted Grammarly’s offer to improve a paragraph. There is no difference from the input I got from editors and beta readers.

I think you’ll understand when you get a rejection. A lot of authors with over a hundred posts have been rejected and don’t know why. I have 110 stores posted. If you say any of them have AI content, then you don’t know what AI is.

A major problem with AI systems is that they cannot explain how they make decisions. The best you can do is test it on additional data and look for patterns. You cannot “program” an AI system to look for things like 'perplexity' and 'burstiness.' All you can do is train it on a lot of data for which you somehow know the truth and look for patterns elsewhere. I find it amusing that people on this thread have added spaces to text and achieved better scores from their favorite AI detector.

Basically, I agree with the posters who called this a witch hunt. Be afraid. Be very afraid. You may be next.

I use google docs. It has a small editor in it but doesn't change entire sentences or paragraphs and allows me ease of use for minors errors. It's one of the better tools imho for amateur writers along with it's share ability for editors.

I dislike grammerly as it loves to change too many things to make it seem human. More uniformly written would be a clue, for me at least, of an AI helper.

 

[Britva415]

Of course, it matters if there is a reliable way to detect AI content.

I was just talking specifically about the practicality of Lit interactions given the rules, not about the global philosophical angles.

As a purely practical matter, saying out loud that you're actually doing the thing which Lit doesn't allow is... probably not going to help you pass approval.

I hope you didn’t mean for your reply to be so condescending and sanctimonious.

I kind of meant for it to be unsympathetic, given the above. I'm sorry, but coming right out and saying you're literally breaking the rule - hey, if you're doing it as some kind of protest move or something, more power to ya, I can respect taking a bullet for everyone if that's what your intentions are, but my respect for being subversive isn't compatible with sympathy for the likely consequences.

I don't know if that's the same as condescending and sanctimonious. I can't control how you hear it, anyway, so, if that's how you hear it, then there you have it. Whether I say "I didn't mean it like that" or not, probably makes no difference to you.

 

[XerXesXu]

A major problem with AI systems is that they cannot explain how they make decisions. The best you can do is test it on additional data and look for patterns. You cannot “program” an AI system to look for things like 'perplexity' and 'burstiness.' All you can do is train it on a lot of data for which you somehow know the truth and look for patterns elsewhere. I find it amusing that people on this thread have added spaces to text and achieved better scores from their favorite AI detector.

Basically, I agree with the posters who called this a witch hunt. Be afraid. Be very afraid. You may be next.

AI detectors aren't trained on the output of the LLM base model. That output's retrained to produce an output for which the training set is human writing trained in the style required by the LLM creators. When you're flagged you're flagged as producing human writing of a certain style. Humans do that, and some will do it with a style similar to eg: the NYT style guide - they claim that Open.AI used the output of their journalists, written in the style of their style guide, to fine-tune ChatGPT - and will be flagged as writing in eg: ChatGPT style.

The site has a choice between doing nothing and doing something. It's chosen to do something with what's available. Think dumb bombs and collateral damage; there will inevitably be collateral damage if the site chooses to do something. I guess very few authors would want it to do nothing.

 

[MillieDynamite]

I don't think your rejection has anything to do with AI. It will at least say "possible AI content" or something to that effect. The story is essentially already up. You can't have the same story in two different postings. Re-posting it in chapter form isn't different enough to qualify as another story. I believe if you want to post it as a series of chapters, you'll need to take the complete story down first. That's just my opinion. You could message Laural and get that straightened out.

I enjoyed them. having them all like that made it great to keep the story together and with chapters 6 and on being separate, not a bad idea as it can keep them coming back and makes the burden of writing a little more tolerable in the long run.

 

[NuclearGorilla]

I think the solution is to try to write less good but more better. There’s no way an AI can figure out how to do that. It’s a purely human skill.