my story keeps getting flagged for AI, NEED HELP

[lil_trip]

I have been working on a series since the past 2 weeks. My first 2 stories got published. Unfortunately the other 3 keep getting sent back. Need help

 

[PennyThompson]

Are you using Grammarly or other writing aid software that offers to tweak or re-write your text? Because those will get flagged as AI.

So will a lot of translation tools, so if you're writing in a language other than English and then running it through a translator, that can also get your story jammed up.

 

[ronde]

My rule of thumb for editing both here and in my professional correspondence is to never, never, ever accept any suggestion from any software program about what I need to change. All are limited to what they "know" and what they "know" is determined by either the dictionary and grammar rules added by some programmer, or in the case of AI, what the developer used as base data. Also, if you have any feature with the label "auto" in the title, disable it.

My favorite example of why you should carefully examine each "suggested change" is one I found in professional correspondence emailed to the immediate world by a young mechanical engineer. There is a term used in the auto industry called "commonize". The definition is to use the same part across several platforms to reduce complexity and cost, like using the same screw in several different places in several different models of cars. This engineer kept writing "commonize" when she typed, but her autocorrect spellchecker always changed it to "Communize." She never verified the change until after the fourth time I read that she was going to "Communize" this part, I told her that I doubted the parts she was talking about were going to take over the means of their production.

 

[YmaOHyd]

I told her that I doubted the parts she was talking about were going to take over the means of their production.

Widgets of the world, unite! You have nothing to lose but your users!

 

[Jorunn]

Definitely take it to Notes or a txt editor and cut and paste from there into Lit to be sure you have no hidden artifacts. Look at long sentences with commas and break them up. Grammarly loves to suggest AI changes I disagree with. Check your dialogue and use contractions - change “I cannot” to “I can’t”. Add slang that AI doesn’t know and unique character names. In dialogue complain about the “stinkin’ AI” instead of the “stinking AI”.

And after none of this works, keep trying!

 

[AwkwardlySet]

I think it would be useful if we created a guide of a sort, something that would include our best guesses and advice on how to approach the problem of AI rejection. It's tiresome to repeat all these thoughts over and over again, yet it also feels wrong to just ignore these people who come here seeking help.

Maybe we can do what Laurel won't?

A step by step guide to resolving (unjustified) AI rejections. It probably wouldn't solve every such rejection but it would be an improvement. It would be something we could offer to rejected authors, something that wouldn't need to include our arguments and snark. Just an idea.

 

[BobbyBrandt]

There are no definitive, proven solutions to be offered. All there are are suggestions that others have found successful in their attempts to avoid or get passed an AI rejection. These include:

Many will tell you not to use any AI-assisted tools, such as Grammarly, and that advice is partially correct. You can use them to identify issues, but always correct the identified issues yourself rather than letting the tool do it for you.

Most AI detection processes report back on the percentage of suspected AI content within the analyzed work. You mention publishing piece-meal work (series), so consider that a report of 50% suspected AI content in a 10K word story could become only 10% in a 50K word story. Maybe submit more of your work content at a time.

Dialogue is often the crucial content that distinguishes human from AI created work. Let your characters speak to your originality.

Change how you submit your work. If you typically paste your content into the submission form, try uploading it instead. If you uploaded it, try pasting it.

I have never had an AI rejection here. I write using MS Word (typically averaging 50K words), always use Grammarly (as described above), and always upload my stories as *.docx files. What works for me isn't guaranteed to work for others, but there it is.

 

[BobbyBrandt]

I think it would be useful if we created a guide of a sort, something that would include our best guesses and advice on how to approach the problem of AI rejection. It's tiresome to repeat all these thoughts over and over again, yet it also feels wrong to just ignore these people who come here seeking help.

Maybe we can do what Laurel won't?

A step by step guide to resolving (unjustified) AI rejections. It probably wouldn't solve every such rejection but it would be an improvement. It would be something we could offer to rejected authors, something that wouldn't need to include our arguments and snark. Just an idea.

I try to do just that, and would be willing to contribute to just such a collaborative effort.

 

[old_prof]

Widgets of the world, unite! You have nothing to lose but your users!

I thought maybe they might worry about losing their nuts. If mine were removable, I'd worry too.

 

[old_prof]

Most AI detection processes report back on the percentage of suspected AI content within the analyzed work. You mention publishing piece-meal work (series), so consider that a report of 50% suspected AI content in a 10K word story could become only 10% in a 50K word story. Maybe submit more of your work content at a time.

You keep repeating this assertion about combining, but it makes no sense. If only one of your five stories was being rejected, maybe. That would argue that you have a very isolated problem. So having the five stories could be an advantage, understanding where that problem lies.

On the other hand, if each of the five stories is roughly 50% AI, than your combined book will stlll be 50% AI.

You only dilute the percentage by combining if your add 0% parts to it. If someone can write overwhleminingly 0% parts, there's probably not a long term problem.

We also have no idea that the system they use reports percentages. The online ones do, but they perform so badly that I question if they're even intended to work. Or simply confuse people about what makes things feel like AI.

 

[BobbyBrandt]

You keep repeating this assertion about combining, but it makes no sense.

I try to only make this "assertion" in response to commenters who state that only some of their submissions are being rejected for suspected AI content, as the OP did in this case.

If they published two 10K word story segments and had no issues but did on the next 10K submission, I will stand by my "assertion" that combining the three parts could possibly yield a different result where the suspected AI content is concerned.

While it is true that we don't know what AI detection process is used here, it is not that much of a stretch of logic to assume that they didn't reinvent the wheel and are employing a process that would also measure a suspected percentage of AI content in the work analyzed.

It's just a suggestion and no one is required to follow it against their will. Lighten up.

 

[EmilyMiller]

I think it would be useful if we created a guide of a sort, something that would include our best guesses and advice on how to approach the problem of AI rejection.

I think that sounds lovely, but I think it’s chimera. Things thag have been suggested:

1. Don’t let Grammarly change your text. This is sound advice as the use of Grammarly in this way is explicitly mentioned in rejection notes. Equally multiple authors use it, but make their own fixes when a problem arises, and they seem to publish fine.
2. Introduce spelling and grammar errors - as GenAI is so perfect. This seems like wishful thinking. And GenAI comes up with clunky grammar too.
3. Add more dialog - my stories are dialog-rich, maybe it helps .
4. If you paste into the text box, try uploading a document instead, or vice versa. This seems like bizarre superstition to me. To my certain knowledge, people have got AI rejections with both approaches. It seems likely to be irrelevant.
5. But… some people have said that Grammarly might leave tokens in your Word file. I have no idea if that’s true, but it would argue to use the text box and not to swap to document uploading. But I’d need to see some more evidence that this is a thing.
6. Avoid em dashes - and look out for black cats crossing your path as well.

The problem with any crowd-sourced document is that it would be entirely based on hunches and guesses. What value would such a document really have?

 

[AwkwardMD]

I think it would be useful if we created a guide of a sort, something that would include our best guesses and advice on how to approach the problem of AI rejection. It's tiresome to repeat all these thoughts over and over again, yet it also feels wrong to just ignore these people who come here seeking help.

I'm surprised no one thought of that yet.

 

[AwkwardlySet]

I'm surprised no one thought of that yet.

If you have something like that in your thread in the editor forum, then this is proof of how few people know about it. I knew you had a thread where you try to help people, but I was thinking something like a self-help guide, something authors could try before asking for your help.

Again, if there's already something like that in your thread there, then maybe we should make a dedicated thread here with links leading to your thread, maybe add some short instructions, etc. It's clear not everyone finds your thread, and you can see that many of us active AHers don't know exactly what it has and what it doesn't have.

 

[AwkwardMD]

If you have something like that in your thread in the editor forum, then this is proof of how few people know about it. I knew you had a thread where you try to help people, but I was thinking something like a self-help guide, something authors could try before asking for your help.

Again, if there's already something like that in your thread there, then maybe we should make a dedicated thread here with links leading to your thread, maybe add some short instructions, etc. It's clear not everyone finds your thread, and you can see that many of us active AHers don't know exactly what it has and what it doesn't have.

Your approach suffers from not actually knowing what works and doesn't. Additionally, you run the risk of giving bad actors more information than they should have.

Mine is, as you said, a place to ask for help and be dealt with on a case by case basis.

 

[AwkwardlySet]

The problem with any crowd-sourced document is that it would be entirely based on hunches and guesses. What value would such a document really have?

I understand all the downsides, but it beats what we do here now. The frustrated posts that rejected authors make either get ignored or get snarky replies, or they get told that their writing likely sucks anyway, or we start arguing in them. It's still an improvement, as lacking as such a guide would be.

 

[EmilyMiller]

I understand all the downsides, but it beats what we do here now. The frustrated posts that rejected authors make either get ignored or get snarky replies, or they get told that their writing likely sucks anyway, or we start arguing in them. It's still an improvement, as lacking as such a guide would be.

I agree that’s a point. We could make a dedicated thread and have people add their experiences. So long as we tag it as definitively not being authoritative and certainly not a universal panacea.

 

[AwkwardlySet]

Your approach suffers from not actually knowing what works and doesn't. Additionally, you run the risk of giving bad actors more information than they should have.

Mine is, as you said, a place to ask for help and be dealt with on a case by case basis.

See the reply to Em. Your thread, as good-intentioned as it is, makes you another Laurel of a sort. You're gatekeeping in the sense of who deserves help and who doesn't. And again, it's AH that gets swarmed with these requests for help, so whatever approach is the best, it should be there as well.

 

[AwkwardMD]

See the reply to Em. Your thread, as good-intentioned as it is, makes you another Laurel of a sort. You're gatekeeping in the sense of who deserves help and who doesn't. And again, it's AH that gets swarmed with these requests for help, so whatever approach is the best, it should be there as well.

Laurel works two cubicles down from me. We're totally different people.

 

[AwkwardMD]

And yes. I am gatekeeping. Getting flagged at all is, ironically, a red flag.

 

[ronde]

I think that sounds lovely, but I think it’s chimera. Things thag have been suggested:

1. Don’t let Grammarly change your text. This is sound advice as the use of Grammarly in this way is explicitly mentioned in rejection notes. Equally multiple authors use it, but make their own fixes when a problem arises, and they seem to publish fine.
2. Introduce spelling and grammar errors - as GenAI is so perfect. This seems like wishful thinking. And GenAI comes up with clunky grammar too.
3. Add more dialog - my stories are dialog-rich, maybe it helps .
4. If you paste into the text box, try uploading a document instead, or vice versa. This seems like bizarre superstition to me. To my certain knowledge, people have got AI rejections with both approaches. It seems likely to be irrelevant.
5. But… some people have said that Grammarly might leave tokens in your Word file. I have no idea if that’s true, but it would argue to use the text box and not to swap to document uploading. But I’d need to see some more evidence that this is a thing.
6. Avoid em dashes - and look out for black cats crossing your path as well.

The problem with any crowd-sourced document is that it would be entirely based on hunches and guesses. What value would such a document really have?

I think you have hit the nail squarely on the head and driven it home with just one blow with your first suggestion.

I've read through most of these, "My story was rejected for AI so what do I do now" threads and one commonality seems to be apparent. The poster will say "I only used (Word, Grammarly, etc.) to check for spelling and grammar." What I've seldom read is how the poster used the software to check for spelling and grammar errors.

I have always used the spellchecker in Word but never use the suggested change without reading it to see if it makes sense to me. Any spellchecker can do some really weird things with words commonly used in erotica. Any spellchecker/grammar checker will usually attempt to change your wording into "business acceptable" text. I constantly get warnings about my use of passive case in my stories and I never change to the suggestion because that changes the story. Another failure that's easy to create is the little box in the spellchecker that says, "Add To Dictionary". Never do this because you'll end up with the spellchecker changing your correct word to a word you've added to the dictionary and the difference will stick out like a wart on a bikini model's ass.

I've cut and pasted my stories into the submission box from Word since my first story. I have yet to have a rejection for AI. Not to say it can't happen, but I have several stories published since the AI rejections seem to have appeared.

 

[PennyThompson]

I've read through most of these, "My story was rejected for AI so what do I do now" threads and one commonality seems to be apparent. The poster will say "I only used (Word, Grammarly, etc.) to check for spelling and grammar." What I've seldom read is how the poster used the software to check for spelling and grammar errors.

I'm thoroughly convinced that this is the source of pretty much every single "false positive" where an author isn't simply lying about using AI.

What Grammarly does when modifying your writing for things like grammar and tone and tenses is mechanically identical to what LLMs do.

Microsoft and Google are both adding more and more LLM slop into their writing tools too, so I expect we'll see more and more of these kinds of triggers.

Write for yourself, don't ever let any tool of any kind write for you!

 

[SkyBubble]

Definitely take it to Notes or a txt editor and cut and paste from there into Lit to be sure you have no hidden artifacts. Look at long sentences with commas and break them up. Grammarly loves to suggest AI changes I disagree with. Check your dialogue and use contractions - change “I cannot” to “I can’t”. Add slang that AI doesn’t know and unique character names. In dialogue complain about the “stinkin’ AI” instead of the “stinking AI”.

And after none of this works, keep trying!

Trying using "gonna' for "going to" when characters are speaking. Human things like that.

 

[EmilyMiller]

Trying using "gonna' for "going to" when characters are speaking. Human things like that.

No!

This is an old wives’ tale.

Written by CharGPT:

—

I was gonna just stay home and stream something, but I kinda missed the whole movie theater vibe, so I grabbed my keys and headed out. The lobby was buzzing, people arguing about what to see, someone spilling popcorn like it was their full-time job. I stood there forever trying to decide between the giant soda I absolutely didn’t need and the slightly smaller giant soda I also didn’t need, and obviously I went big. The trailers were low-key louder than a rocket launch, and everyone was half on their phones but still pretending they weren’t. Once the lights dimmed, though, it was actually pretty awesome—like, there’s just something about that massive screen and surround sound that hits different. By the time I walked out, blinking into the parking lot lights, I was already kinda planning what I’m gonna see next weekend.

 

[Jorunn]

No!

This is an old wives’ tale.

Written by CharGPT:

—

I was gonna just stay home and stream something, but I kinda missed the whole movie theater vibe, so I grabbed my keys and headed out. The lobby was buzzing, people arguing about what to see, someone spilling popcorn like it was their full-time job. I stood there forever trying to decide between the giant soda I absolutely didn’t need and the slightly smaller giant soda I also didn’t need, and obviously I went big. The trailers were low-key louder than a rocket launch, and everyone was half on their phones but still pretending they weren’t. Once the lights dimmed, though, it was actually pretty awesome—like, there’s just something about that massive screen and surround sound that hits different. By the time I walked out, blinking into the parking lot lights, I was already kinda planning what I’m gonna see next weekend.

None of this is conversation. I would use gonna as part of a conversation. Also, shorter paragraphs. And mix text with conversation in the same paragraph.

I turned my head to look at the sudden noise. It was Karl! I yelled, “Haloo, old friend!”