Why does Lit keep rejecting my story?

[MsPeachPit]

Several months ago I published a story on Lit with no problem whatsoever. I took it down six weeks ago to remove a few paragraphs that I didn't like and then tried to republish the story. Lit returned it, claiming that there was evidence of AI used in the writing. The fact is that I wrote this story on Word 7 and didn't even use grammar check on it. No other programs were involved at all.So I reworked a few things in the story and resubmitted it with a note that I was using only Word 7. Well, a month later the story was again returned with the same reason for rejection.

Anyone have any idea of what might be going on here? It's very frustrating. BTW, I published two other stories around the same time written on the same Word 7 program and had no problem.

 

[Actingup]

Taking it down was your first mistake!

It must be falling foul of a false positive from AI checkers. I support LitErotica's efforts to keep AI off the site, but it's frustrating to be rejected on that account.

Suggestions:
- search the many AI-related threads on this site for clues.
- put the story through some AI-checkers and see if you can spot anything that they're flagging - rework the story further to ensure that it's in your distinctive author's voice
- give it up as a bad job and write some more stories.

 

[ShelbyDawn57]

Taking it down was your first mistake!

It must be falling foul of a false positive from AI checkers. I support LitErotica's efforts to keep AI off the site, but it's frustrating to be rejected on that account.

Suggestions:
- search the many AI-related threads on this site for clues.
- put the story through some AI-checkers and see if you can spot anything that they're flagging - rework the story further to ensure that it's in your distinctive author's voice
- give it up as a bad job and write some more stories.

Write some more stories regardless...

 

[Five_Inch_Heels]

Several months ago I published a story on Lit with no problem whatsoever. I took it down six weeks ago to remove a few paragraphs that I didn't like and then tried to republish the story. Lit returned it, claiming that there was evidence of AI used in the writing. The fact is that I wrote this story on Word 7 and didn't even use grammar check on it. No other programs were involved at all.So I reworked a few things in the story and resubmitted it with a note that I was using only Word 7. Well, a month later the story was again returned with the same reason for rejection.

Anyone have any idea of what might be going on here? It's very frustrating. BTW, I published two other stories around the same time written on the same Word 7 program and had no problem.

Is English your first language? The one you grew up and use all the time?



The bot checker can fart on stuff written by people who learned English in school later in life.

 

[AwkwardMD]

Is English your first language? The one you grew up and use all the time?



The bot checker can fart on stuff written by people who learned English in school later in life.

I would like it on the record that I do not think this to be true. Lit's AI Detector is looking for specific things, and I don’t think this is one of them.

That being said, I also  believe that some or all flagged stories then get a cursory lookover from the site moderator, and it is entirely possible that a non-native writing would be more likely to fail the eye test. In other words, it's not what got you flagged in the first place, but the chances that it hurts you if you get to this step are non-zero.

(For the purposes of this post, I am drawing a distinction between think and believe where think means something I am like 90% sure of and believe is closer to 50%)

 

[Bamagan]

Several months ago I published a story on Lit with no problem whatsoever. I took it down six weeks ago to remove a few paragraphs that I didn't like and then tried to republish the story. Lit returned it, claiming that there was evidence of AI used in the writing. The fact is that I wrote this story on Word 7 and didn't even use grammar check on it. No other programs were involved at all.So I reworked a few things in the story and resubmitted it with a note that I was using only Word 7. Well, a month later the story was again returned with the same reason for rejection.

Anyone have any idea of what might be going on here? It's very frustrating. BTW, I published two other stories around the same time written on the same Word 7 program and had no problem.

Maybe yes, but only an idea. Whatever Lit is using as a detector, all it can probably be doing is looking at sentences and paragraphs and comparing them to the 'ideal' output of one or more LLMs. There are a number of features that are suggestive of a formulaic approach to writing, such as the neutral or generic tone it usually adopts (unless prompted to mock some famous writer's style), and either overreliance on or conspicuous avoidance of certain words and phrases. The presence of such constructions in human writing is unavoidable, so I presume the site is either looking at the prevalence/ratio of suspicious content, and/or looking for sections that display significant differences in style from each other, which may indicate copying from an AI (or perhaps direct plagiarism).
Since your second revision was pending for a month and still rejected, that may be the end of the road, at least for your current version. It suggests (to me) that it was sitting on a to-do pile until Laurel had the time to actually read at least a few paragraphs or sections instead of skimming them.
One piece of advice I think is bad to follow is the suggestion that you need to add errors to avoid rejection. For one thing, while LLMs don't usually make spelling errors, they are still prone to other kinds of mistake, such as confusing homonyms despite clear context clues. Including grammatical constructions that would be considered incorrect (or at least inappropriate) in business or academic writing is probably a good idea, though. The modern LLMs incorporate basically the same 'grammar matrix' that has been available for years somewhere under the hood, so following the advice of even older versions of Word too closely can result in sentences that read or feel basically the same as what a modern bot would construct, given the right prompts.
Good luck.

 

[Kelliezgirl]

Send a note to Laurel explaining what happened.
It's frustrating, and there are certainly some challenges with the AI detection, but at this point some form of anti- AI is a necessary evil.
Don't give up!

 

[MsPeachPit]

Maybe yes, but only an idea. Whatever Lit is using as a detector, all it can probably be doing is looking at sentences and paragraphs and comparing them to the 'ideal' output of one or more LLMs. There are a number of features that are suggestive of a formulaic approach to writing, such as the neutral or generic tone it usually adopts (unless prompted to mock some famous writer's style), and either overreliance on or conspicuous avoidance of certain words and phrases. The presence of such constructions in human writing is unavoidable, so I presume the site is either looking at the prevalence/ratio of suspicious content, and/or looking for sections that display significant differences in style from each other, which may indicate copying from an AI (or perhaps direct plagiarism).
Since your second revision was pending for a month and still rejected, that may be the end of the road, at least for your current version. It suggests (to me) that it was sitting on a to-do pile until Laurel had the time to actually read at least a few paragraphs or sections instead of skimming them.
One piece of advice I think is bad to follow is the suggestion that you need to add errors to avoid rejection. For one thing, while LLMs don't usually make spelling errors, they are still prone to other kinds of mistake, such as confusing homonyms despite clear context clues. Including grammatical constructions that would be considered incorrect (or at least inappropriate) in business or academic writing is probably a good idea, though. The modern LLMs incorporate basically the same 'grammar matrix' that has been available for years somewhere under the hood, so following the advice of even older versions of Word too closely can result in sentences that read or feel basically the same as what a modern bot would construct, given the right prompts.
Good luck.

Thank you for this. It was very informative and I really appreciate your taking the time to share this.

 

[M_K_Babalon]

I'd have left it alone.

 

[BobbyBrandt]

I would like it on the record that I do not think this to be true. Lit's AI Detector is looking for specific things, and I don’t think this is one of them.

I took @Five Inch Nails comment to imply that maybe there was some application translating the story from some other language to English that could be getting it flagged as AI.

I could be wrong

 

[AwkwardMD]

I took @Five Inch Nails comment to imply that maybe there was some application translating the story from some other language to English that could be getting it flagged as AI.

I could be wrong

Translation apps are, for Literotica's purposes, generative, and would likely trigger a red flag.

 

[Five_Inch_Heels]

I took @Five Inch Nails comment to imply that maybe there was some application translating the story from some other language to English that could be getting it flagged as AI.

I could be wrong

A little of both.

I've seen ESL writers use stiffly structured wording and formatting that native speakers and writers do not. Paragraphs can be complicated and difficult to read at a normal conversational pace. It isn't even always in stories. I've seen it in forum posts where you have to read things several times to understand what they're trying to say.

Quite believable that a bot detector might be triggered by that kind of writing.

 

[MillieDynamite]

Because "belief" isn't based on evidence so much as feelings or faith. And faith is the ability to believe in something without proof.

(For the purposes of this post, I am drawing a distinction between think and believe where think means something I am like 90% sure of and believe is closer to 50%)

 

[AwkwardMD]

Because "belief" isn't based on evidence so much as feelings or faith. And faith is the ability to believe in something without proof.

I have some evidence, but it’s anecdotal and inferred. It's much less solid than my understanding of the base system.

EDIT:let's be real, also some straight up wishful thinking.

 

[MillieDynamite]

Yeah, I got that. I'm just pointing out the difference between believing something, thinking something to be true, and, by mental extension, knowing. Knowing is the hard part; to know something is true requires a lot of evidence. I used to believe a specific, unnamed, public figure was a bit unbalanced, then I thought he was on the edge of insanity, now I know he's beyond dancing cat crazy. The thread probably didn't need my expansion on your statement, but, as everyone knows, not believes or thinks, I go over the top often.

I have some evidence, but it’s anecdotal and inferred. It's much less solid than my understanding of the base system.

 

[SkyBubble]

Maybe yes, but only an idea. Whatever Lit is using as a detector, all it can probably be doing is looking at sentences and paragraphs and comparing them to the 'ideal' output of one or more LLMs. There are a number of features that are suggestive of a formulaic approach to writing, such as the neutral or generic tone it usually adopts (unless prompted to mock some famous writer's style), and either overreliance on or conspicuous avoidance of certain words and phrases. The presence of such constructions in human writing is unavoidable, so I presume the site is either looking at the prevalence/ratio of suspicious content, and/or looking for sections that display significant differences in style from each other, which may indicate copying from an AI (or perhaps direct plagiarism).
Since your second revision was pending for a month and still rejected, that may be the end of the road, at least for your current version. It suggests (to me) that it was sitting on a to-do pile until Laurel had the time to actually read at least a few paragraphs or sections instead of skimming them.
One piece of advice I think is bad to follow is the suggestion that you need to add errors to avoid rejection. For one thing, while LLMs don't usually make spelling errors, they are still prone to other kinds of mistake, such as confusing homonyms despite clear context clues. Including grammatical constructions that would be considered incorrect (or at least inappropriate) in business or academic writing is probably a good idea, though. The modern LLMs incorporate basically the same 'grammar matrix' that has been available for years somewhere under the hood, so following the advice of even older versions of Word too closely can result in sentences that read or feel basically the same as what a modern bot would construct, given the right prompts.
Good luck.

Vary sentence length and structure and paragraph length. That increases what they call "burstiness." You want high burstiness. LLMs tend to produce low burstiness and the AI checkers tend to see that as a sign of AI use.

 

[MillieDynamite]

Other than dialog, do not have many one-line or sentence paragraphs in a row. I don't know if this is an AI rejection thing or not, but it isn't good writing, and sounds like AI to me. Your writing should be a rollercoaster, with long paragraphs, followed by medium, followed by very short. Sentence structure should be the same. Varying your length in both makes reading easier. It also keeps interest up as they go. That's something AI doesn't do, it's passive writing, and passive writing is boring. AI checkers, unlike AI writers, know boring!

 

[StillStunned]

Other than dialog, do not have many one-line or sentence paragraphs in a row. I don't know if this is an AI rejection thing or not, but it isn't good writing, and sounds like AI to me. Your writing should be a rollercoaster, with long paragraphs, followed by medium, followed by very short. Sentence structure should be the same. Varying your length in both makes reading easier. It also keeps interest up as they go. That's something AI doesn't do, it's passive writing, and passive writing is boring. AI checkers, unlike AI writers, know boring!

I think your tips about paragraph length need to be qualified a little.

Nowadays you have to consider that a large part of the readership only read on their phones. They don't want endless walls of text to scroll through - they'll be more likely to just skip ahead. Paragraphs nowadays are just shorter than they used to be. My personal guideline is 60-90 words, and often even shorter.

There's some leeway, I think, with stories that offer more emotional engagement. A longer paragraph does give you the opportunity to dig into feelings without breaking the train of thought. But even then, you have to have your reader hooked and willing to focus that long.

For stories that focus more on action, I'd strongly advise against long paragraphs. Shorter paragraphs keep the story moving along at a rapid pace, at least if you stick to the principle that a paragraph should contain all and only elements that belong together.

I agree entirely with what you say about sentences though.

 

[iwatchus]

So do we have to make our writing worse so it works better on a phone?

EDITTED: And get off my lawn, while you're at it.

 

[StillStunned]

So do we have to make our writing worse so it works better on a phone?

EDITTED: And get off my lawn, while you're at it.

Aww, but the grass is greener over there!

And no, I'm not saying we have to make our writing worse. Personally, I think shorter paragraphs make writing better. Longer paragraphs can quickly become an indulgence of the writer, adding too much information that the reader will probably skip anyway. Back in the day of paperbacks I'd often come across paragraphs that covered a whole page or more. And that's fine with a physical book, with pages that you turn.

But nowadays, the fact is that most people read online, and a huge proportion do it on their phone. If you make your paragraphs too long, you risk the reader zoning out, skipping ahead or just clicking back.

It's a fact of life, and it requires a different approach. Like I mentioned above, I think it works very well to keep the story moving forward, and when you reach the point where you know your reader is engaged you can use longer paragraphs to describe deeper emotions and experiences.

 

[MillieDynamite]

No, nothing of what I'm talking about makes writing worse. Your writing needs to be visually stimulating, as well as the words making it interesting, which is what keeps readers engaged. And I don't go into other peoples sandboxes so don't yell at me about something I haven't done.

So do we have to make our writing worse so it works better on a phone?

EDITTED: And get off my lawn, while you're at it.

 

[iwatchus]

No, nothing of what I'm talking about makes writing worse. Your writing needs to be visually stimulating, as well as the words making it interesting, which is what keeps readers engaged. And I don't go into other peoples sandboxes so don't yell at me about something I haven't done.

My apologies if you thought I was criticizing your points. No slight was intended.

I was actually picking on @StillStunned comment about paragraph length and phones. If shorter paragraphs are better, do them. If not, don't, regardless of the phone readers. Except maybe there is an argument to give one more forcing function for writers to rein in their own indulgences. (What we can be self-indulgent in our writing? Never.)

 

[MillieDynamite]

I wouldn't have used the sticky frog tongue, if I was seriously offended, and I wasn't, and I knew what you meant. But I didn't want my point obliterated by @StillStunned's opinion about what phone readers want or need. I don't write for the device but the experience, and I think one must consider the online desktop and laptop as the primary means of delivery.

My apologies if you thought I was criticizing your points. No slight was intended.

I was actually picking on @StillStunned comment about paragraph length and phones. If shorter paragraphs are better, do them. If not, don't, regardless of the phone readers. Except maybe there is an argument to give one more forcing function for writers to rein in their own indulgences. (What we can be self-indulgent in our writing? Never.)

 

[iwatchus]

I wouldn't have used the sticky frog tongue. I was seriously offended, and I wasn't, and I knew what you meant. But I didn't want my point obliterated by @StillStunned's opinion about what phone readers want or need. I don't write for the device but the experience, and I think one must consider the online desktop and laptop as the primary means of delivery.

My eyes aren't very good anymore, so I often can't recognize the emoji's people use. Besides I was already old when the whole idea of emoji was invented. I never really moved past emoticons (the text version), but I was friends with the guy who invented them (Scott Fahlman), so I can claim sentimental value there. Rather than just being an old dog.

 

[gunhilltrain]

Taking it down was your first mistake!

It must be falling foul of a false positive from AI checkers. I support LitErotica's efforts to keep AI off the site, but it's frustrating to be rejected on that account.

Suggestions:
- search the many AI-related threads on this site for clues.
- put the story through some AI-checkers and see if you can spot anything that they're flagging - rework the story further to ensure that it's in your distinctive author's voice
- give it up as a bad job and write some more stories.

You don't have to remove a story before submitting the new version. Of course, the way Lit works, it's best to have a "final cut" file of a story so you have something to work on. Maybe you'll never need it, but it should be in a file titled, say, "submitted" in case it's ever needed. If the changes are really big, you can just submit a new version which will replace the old one. Do you (@MsPeachPit) get what I'm talking about?