Advice appreciated on how to deal with editor's allegation of AI use

[EmilyMiller]

I make no attempt to refute the madness claim, but I did encounter this problem as well once last year, with a story being rejected. It was sorted out without too much hassle, but nevertheless, it did sting for a while. As a foreigner, it truly is challenging to figure out which words work as a synonym for others, and I think that sometimes WordHippo doesn't offer the best of suggestions.

Furthermore, the sentence structuring is entirely different between Swedish and English - and sometimes that means your mind plays tricks on you. Things look right to me sometimes that is all backwards to a native speaker. Some if it I manage to catch when using text-to-speech during the editing process, as I can tell it doesn't sound quite right, but my English still needs work. That being said, I do feel pride over the fact that it has improved since I started writing here in the summer. Because I believe it has. At least, I second-guess myself less now than I used to.

My only language is basic French (well a bit of Latin, does that count?). I could no more write a story in French than I could fly.

Emily

 

[EmilyMiller]

Trust me. You're not alone. I have begun to question if I was a real person or if I am an AI

Record yourself singing Daisy Daisy. Play it back slooowwllyyy. How do you sound?

Emily

 

[Devinter]

Maybe we really are AI...

... Aren't you using the Linux logo as your profile picture? And you know what? From Wikipedia: At the software level, within an operating system, a port is a logical construct that identifies a specific process or a type of network service.

Coincidence? I think not. Time to go Ex Machina.

My only language is basic French (well a bit of Latin, does that count?). I could no more write a story in French than I could fly.

Emily

... Bet you've flown to heaven once or twice.

 

[storyteller92]

Record yourself singing Daisy Daisy. Play it back slooowwllyyy. How do you sound?

Emily

The best I can do is... Baby shark doo doo doo doo doo baby shark doo doo dooo doo doo doo

 

[Deleted member 6736302]

You can always tell when people start getting their stories posted through the system when they disappear from these threads and I become the lone soul yelling into the void on day 54 since I first submitted this part back in January

One day the void will yell back with something other than 'Sent Back'.

One day.

On the other hand though, rewriting realms of work to get through an unknown set of criteria for the site isn’t much fun. I’ve had to take a break from writing and posting here because I feel burned out from all of it!

 

[Deleted member 6736302]

A raw (bad) conversation passed through an AI to write the actual story. Obviously. Sequences make no sense.

Latest comment I have received on my latest story.

Like, it’s not true - and my god, am I pissed after the two months spent writing and rewriting this.

Fundamentally fuming.

Then there’s this one on an earlier one:

AI drivel. Especially obvious when the "author" tries to write the explicit parts themselves:
"Her pussy was pretty tight, and he was going pretty deep inside her."
So, I guess. Maybe I should be happy an AI wrote the other 90%

So not only have had I the ignominy of being accused of using AI, when I haven’t, but now there’s an element of defending myself from AI allegations from anonymous members.

So frustrating.

 

[NuclearGorilla]

Nothing particularly new to add, but I decided to run the first few pages of one of my pending stories through several AI detection tools and got the following:

QuillBot: 21% likely to be AI
Copyleaks: Definitely AI
GPTZero: 0% to 10% AI depending on the section scanned

This would seem incriminating if I had not developed the story free of AI assistance, thus being reasonably sure of my innocence.

By removing some commas, nicknames, and adjectives at random, I was able to get:

QuillBot: 3% AI
Copyleaks: Definitely AI
GPTZero: 0% AI

This left the paragraphs bland with a mixed reduction in AI false flagging. The only assistance used for these sections was MS Word default spellcheck or Google Docs default spellcheck to scan for typos and comma issues.

Just adding data to the ongoing discussion.

 

[EmilyMiller]

Nothing particularly new to add, but I decided to run the first few pages of one of my pending stories through several AI detection tools and got the following:

QuillBot: 21% likely to be AI
Copyleaks: Definitely AI
GPTZero: 0% to 10% AI depending on the section scanned

This would seem incriminating if I had not developed the story free of AI assistance, thus being reasonably sure of my innocence.

By removing some commas, nicknames, and adjectives at random, I was able to get:

QuillBot: 3% AI
Copyleaks: Definitely AI
GPTZero: 0% AI

This left the paragraphs bland with a mixed reduction in AI false flagging. The only assistance used for these sections was MS Word default spellcheck or Google Docs default spellcheck to scan for typos and comma issues.

Just adding data to the ongoing discussion.

Maybe you did before, but could you post a section that is flagged as “bad” here?

Emily

 

[trevrepa]

I have two stories pending, both have been rejected multiple times and resubmitted. I ran both stories through 10 detectors and here are the results.

Tricked Pt 2 – (5413 words, 30,874 characters)

Quillbot - 2% (first 1200 words)
GTPZero – 1% (first 5000 characters)
Copyleaks – AI Content detected (but only 2 paragraphs highlighted in red for first 25,000 characters)
ZeroGPT – 2.98% (first 15,000 characters)
ContentatScale – Passes as Human
Sapling – 93.5% (first 8000 characters)
Writer – 8% (first 5000 words)
Crossplag – 0% (first 3000 words)
ContentDetector – 31.19% (first 2500 characters)
UndetectableAI – Appears human (entire story)

It passed 8 of 10. Sapling really hates me.


Diary of a Supermodel (2912 words, 16,089 characters)

Quillbot - 21% (first 1200 words)
GTPZero – 0% (first 5000 characters)
Copyleaks – Human text (entire story)
ZeroGPT – 10.07% (first 15,000 characters)
ContentatScale – Passes as Human (entire story)
Sapling – 82.6% AI (first 8000 characters)
Writer – 0% (entire story)
Crossplag – 0% (entire story)
ContentDetector – 47.14% (first 2500 characters)
UndetectableAI – Appears human (entire story)

This story passed 9 of 10 with again Sapling being the lone detector that it didn't pass.

I really don't know what Sapling has against my writing style. I purchased a 30 day trial with them to look into it and found some glaring problems with their engine which I discussed with their support team. I'll post that another time.

 

[lustychimera]

By removing some commas, nicknames, and adjectives at random

That sucks. I put a lot of thought into those and their placement. That stuff's critical to the tone and pacing. Em dash maybe?

 

[DrPope]

By removing some commas, nicknames, and adjectives at random, I was able to get:

QuillBot: 3% AI
Copyleaks: Definitely AI
GPTZero: 0% AI

This left the paragraphs bland with a mixed reduction in AI false flagging. The only assistance used for these sections was MS Word default spellcheck or Google Docs default spellcheck to scan for typos and comma issues.

Just adding data to the ongoing discussion.

The tone and vibe is exactly why your writing was so good ...don't change anything for the AI Witch Hunt Lynch mob ...PM Laural or whoever it is and complain ...then go where I told you via email.

If your having AI trouble anyone who is good at writing in that informal style is in trouble ....You write some of the most original prose out there....yes it's in the direction of DBBQ but it's very different in other ways of actual writing style and excellent at that.

Honestly it's just a terrible state of affairs here now .... why there isn't a massive white list of authors by now submitting unfettered is frankly beyond my comprehension ...it's a no brainer to only check new authors works for AI ...

 

[Deleted member 6736302]

And there we go.

After 2 weeks, I've got the rejection note I thought I was going to get a week ago.

I can't do anything else.

I'm done.

This is shit, to be honest.

I’ve read your work, there was nothing AI about it.

Really is poor form given the number of PMs, threads, posts and more that there is no official line other than that in the rejection sections.

 

[DrPope]

but she said that my work is still coming back as AI-generated and I just don't get how

What she means is "My AI detection is shit and I don't know what to do"

This is now becoming victim blaming .... its offensive to authors

 

[trevrepa]

I did receive a PM from Laurel - thanks to Emily reaching out about the issue - but she said that my work is still coming back as AI-generated and I just don't get how

Did she at least highlight where in the story so you can work on that section? Otherwise it's like trying to find a needle in a haystack.

 

[toughfire]

View attachment 2329148

 

[StillStunned]

All that tells you is that the Harry Potter books have been used extensively to feed AI.

 

[storyteller92]

View attachment 2329148

Just wait until you try poetry or rhyme and verse XD

I have two stories pending, both have been rejected multiple times and resubmitted. I ran both stories through 10 detectors and here are the results.

Tricked Pt 2 – (5413 words, 30,874 characters)

Quillbot - 2% (first 1200 words)
GTPZero – 1% (first 5000 characters)
Copyleaks – AI Content detected (but only 2 paragraphs highlighted in red for first 25,000 characters)
ZeroGPT – 2.98% (first 15,000 characters)
ContentatScale – Passes as Human
Sapling – 93.5% (first 8000 characters)
Writer – 8% (first 5000 words)
Crossplag – 0% (first 3000 words)
ContentDetector – 31.19% (first 2500 characters)
UndetectableAI – Appears human (entire story)

It passed 8 of 10. Sapling really hates me.


Diary of a Supermodel (2912 words, 16,089 characters)

Quillbot - 21% (first 1200 words)
GTPZero – 0% (first 5000 characters)
Copyleaks – Human text (entire story)
ZeroGPT – 10.07% (first 15,000 characters)
ContentatScale – Passes as Human (entire story)
Sapling – 82.6% AI (first 8000 characters)
Writer – 0% (entire story)
Crossplag – 0% (entire story)
ContentDetector – 47.14% (first 2500 characters)
UndetectableAI – Appears human (entire story)

This story passed 9 of 10 with again Sapling being the lone detector that it didn't pass.

I really don't know what Sapling has against my writing style. I purchased a 30 day trial with them to look into it and found some glaring problems with their engine which I discussed with their support team. I'll post that another time.

I have been testing, PRIMARILY on Quillbot and checking Zero GPT and a few others 500-1200 words at a time.
ALL of my writing averages less than 20 percent. (I only say twenty because one passage came back at 17%, and I don't want to be dishonest. I wrote TWO stories offline, hit F7. Both were about 6000 words long. One averaged (unweighted) 3.5% with one detector claiming there was 8%, which was the highest I could find against it.
The stories were shot back to me.
I get why Portly is upset. I write here, trying to get feedback and improve, but also because I don't think my idea/stories are half bad, but when we are denied before we start, it can be disheartening.
Telling us to leave in grammatical errors isn't the answer because then it's just saying leave the work unpolished. If that's the solution, why even have an editors' section?

 

[toughfire]

All that tells you is that the Harry Potter books have been used extensively to feed AI.

I see the mentality behind it.

Truly sad to see talented authors being punished for nothing, though.

 

[RejectReality]

All that tells you is that the Harry Potter books have been used extensively to feed AI.

What it tells you is that they're cash-grab garbage that don't perform the service they're advertising. When something tells you that Harry Potter and the Constitution are written by AI, it has zero credibility.

For everyone suggesting there should be itemized lists of what's triggering the rejections...

Why?

You want to change your voice to appease some shitty software that fails so spectacularly? What do you do when said shitty software gets an update and it starts flagging something else?

One rejection, one resubmission with no changes, and then take it somewhere else if it's rejected again. Tilting at windmills is stifling you. Every moment you spend fighting a brainless AI monster is a moment you could be working on your next story. The audience may not be as big as Lit's, but at least it will get read.

 

[toughfire]

It's interesting to mention that even the Sapling detector warns that false positives and false negatives will occur. It's a scary time to be a writer.

I imagine this process will influence people to come up with very avant-garde and clunky prose, in order to pass the test. As a fan of simple prose, that sounds like a nightmare to me.

 

[storyteller92]

I see the mentality behind it.

Truly sad to see talented authors being punished for nothing, though.

Honestly, if the author is less than a certain threshold and has been publishing somewhat consistently, there needs to be a white/gray list.
Also, some transparency on percentages might prove helpful. I don't have all the answers but this making authors wait will become untenable.

 

[StillStunned]

I imagine this process will influence people to come up with very avant-garde and clunky prose, in order to pass the test. As a fan of simple prose, that sounds like a nightmare to me.

My prose is probably the simplest you can imagine, and (so far) none of my 28 stories have been rejected for being AI.

 

[toughfire]

Honestly, if the author is less than a certain threshold and has been publishing somewhat consistently, there needs to be a white/gray list.
Also, some transparency on percentages might prove helpful. I don't have all the answers but this making authors wait will become untenable.

I imagine most familiar authors on the forum are not being rejected. But there are several cases of authors who have published tons of stories for a long period of time, and were suddenly flagged. In my case, I had been publishing for about a year until one of my stories was flagged.

 

[Deleted member 6736302]

I did receive a PM from Laurel - thanks to Emily reaching out about the issue - but she said that my work is still coming back as AI-generated and I just don't get how

Absolutely bonkers. Has she physically read the work?

I can’t fathom how it comes back as AI. I just can’t.

 

[Deleted member 6736302]

Just as a side note - I was getting tons of rejections as @Portly_Penguin is aware. I have had a total of 31 AI rejections over February and start of March.

Here’s what I have done for myself to get through:

• No stories over 4000 words
• Check of dialogue to forum approved standard
• No use of grammarly, which has been my go to for some time
• All stories now written in apple notes and copy/pasted into the forum box
• All non essential descriptive sentences removed, these came up as AI derived most often
• Making sure I say he/his/she/her plus names as a deliberate mixture in paragraphs
• No excessive dialogue - over long sentences

It’s been a lot of changes and work to make sure the writing is good enough to get through.

I have physically rewritten chapters for Lit specifically to get through and whilst the story is the same, the actual form and wording isn’t what I originally wrote by some way.

Yes, you can make it work. Is it a bridge too far? For me, I want to see how the story is received and the writing is secondary to that, but others will differ.

I am gutted for @Portly_Penguin and I hope this gets sorted for him soon.