#case: closed

[bookmarkburn]

You said come back with the name. And I get that ”crybaby” makes sense if you know but WE don’t know, so can we keep it civil please. It’s The Girl with the Butterfly tattoo. Your bookmarks said nothing useful and she’s made it sound like you’re out to get her or something for no reason than to fuck with her. She’s been telling everyone how she writes!’ So like, what do you actually have to go on? Besides that it’s a little off, all of it. I see that. But off doesn’t really mean anything, people have different styles.

I don’t mean to be rude but try to see it from my side. It’s semi distressing to think you’re just trolling and stressing her out for nothing OR she’s got everyone believing her but she’s actually the one fucking with us. Sorry but you stirred shit, so can you just tell us?

”Oh for Pete’s sake”, because I thought you meant someone else I gave two heads-ups for.

I’ll be civil. I’ve sure got something to go on. Get comfortable, there’ll be a full-length Heart Starts Pounding Podcast below the cut. Contrary to popular belief though, I don’t enjoy this. It’s ridiculous.

(Anon first submitted this)

Once more for the cheap seats in the back: AI is allowed on AO3. I have no beef with that, I have beef only with the people trying to cheat others into believing they’ve written the fics themselves. Beef isn’t fun. This isn’t fun. But realizing you’re reading AI slop is not fun either.

I’m not out to “get her”, as you say, Anon. I’ve never spoken to her minus commenting. I couldn’t care less who she is. And I didn’t go looking for users to burn, this one was handed to me. On Monday 25th August, I checked my texts. There was a lot to read in the following days. So it’s not like I didn’t have time to make sure that when I commented, case was airtight. I didn’t bookmark until two weeks later. Comment moderated, no response, blocked. Nothing changed in tags or author’s notes. Then my partner commented too to tag the work. Blocked. Then more works showed up, still no mention of AI use.

That’s when we colored me pissed off. Hence the second bookmark’s tone. Not about her as a person, about her actions:

1) Continuing to lie (on a Presidential Campaign level, if what you say is true), and

2) continuing to smuggle AI where it’s explicitly not welcome (Fests)

But I seriously hadn’t even thought of her again until you dropped in, I’ve got a lot of these going on. But here you are, so here I go. I wonder though why she’s not in my asks or DMs demanding evidence based on what you say here and in the previous ask, if it’s been that ”loud”? I’ve received nothing.

But then I think about it and, in her shoes, I probably wouldn’t either, when this is what it is.

This one’s a relatively new author, first work ever published November 2024 (multi-chapters from December). She could have responded “Sorry, I didn’t actually know it was that big of a deal, I’ll add a note.” I mean, she’s gifted a shitload of AI works to people, so presumably there’s a friend group. Could have said “Guys, sorry, I didn’t realize I should inform people…” Nothing. Nothing, because she knows what she’s doing is crap.

Do I need to remind some people why fanfiction writers hate AI? How fundamentally violating it is? Why we don’t want to get tricked into reading it? Because it is tricking people into thinking you’re a brilliant writer, rolling in their love, and then doing it again. And again. +10 works of hers are in Fest collections. What do these collections say in the rules? 

Absolutely No AI Works (Wake Me Up When September Ends Fest) 

NO use of AI will be permitted (Summertime Sadness Fest) 

We do not support or allow the use of AI to create stories or art.(HP Daddy Knows Best Fest) 

And the list goes on. Can these moderators make it any clearer that AI content is not welcome? No. But some people just don’t care or respect that. Because if no one knows, no one gets hurt. Right? 

I realize so far this is bark little bite. Bear with me. You’re tired of her? I find this in such bad taste I can’t even, and I wasn’t gonna give this one another minute of my time. But here we go:

You said she says she writes all her own work. I don’t doubt that she writes. Might even write herself and then feed it to AI that transforms it, she’s clearly got ideas that need to be put to paper or this wouldn’t be happening. Creative urges are incredibly strong and they should be realized. But not through lying, not through gaining the trust and admiration of people who are vehemently against AI.

I understand that you, Anon, might not see what I (and others) see, so I’m going to provide evidence. I hate the hatebots, I hate baseless accusations too. But nothing about this is baseless. I actually don’t care if anyone believes me, hence why I’m not obsessively posting usernames or Blazing to get their asses. You can’t Blaze my posts.

I just don’t want to read slop when I’m on AO3 trying to find cool stories. I can’t filter out things that people don’t tag. And bookmarking is for READERS. And because it’s for readers, I will update the information there soon, as requested, so that people can see I’m not just making shit up for the hell of it. 

Onto the evidence. 

I’m not gonna provide you with a full list. Why? 

1) This user’s works are (minus first One Shot and two earliest MCs) so identical in style and cadence and arrangement and word order and structure and formulation (yes, long list) that what is said here applies overall, and Tumblr posts have word limits.

2) It’s not only people who hate AI who follow this blog. It’s also people looking for ways to slip under the radar. I’m AI users will run and delete all trace of every single thing I mention here. Go right ahead, it won’t make enough of a difference. I have dated copies of not just the bookmarks but other works of the same users downloaded before commenting + stuff I’m sitting on. Because people really are that fucking predicable.

First Point: Volume of Produce. 

Since the first of November 2024, this author has published 598,964 words.  

Let’s just take a moment to consider that number. Not even a year. No mention of backlog in MC opening notes. Sure, people can be unemployed, summer break. Many things. But for reference, Manacled was around 370,000. The longest HP around 257,000. The LotR trilogy around 481,000.

Over half a million words. About 61K a month. Is this impossible? No! But it’s over half a million words of perfect cadence and perfectly smooth, perfectly consistent style. Coincidentally: the perfect cadence and perfectly smooth, perfectly consistent style of style-directed ChatGPT. The only visible change in any of this is how in spring the style seemed to suddenly lock in (earlier works still read generated but with a more default narrative style) and how at one point, this user swapped between m-dash alignment and then reverted between works. That’s just an observation. So I’ll let you consider that whooping word count for yourself. 

Second Point: Randomized Examples & Overlap

Randomized as in a random work that’s part of a No AI Allowed-collection. But this doesn’t really matter because as I said, from this spring onward, it’s all that consistent. You’ll pretty much find what I observe here whichever One Shot you click on from early this year.

First Example: In Nomine Patris et Daddii. What do I see? Or more like, what did I recognize that I’ve seen just a little too much of on AO3 since last fall? And in manuscripts we’ve rejected? And before you start screaming at me: 

I never need to put an author’s work into AI detection software or other AI, and I don’t advise people to do this.But you need hard facts, and I can’t feel bad because I know for a fact that I’ve only fed it back its own slop for dinner. But just this once. No other works.

So, what did I actually do? 

I wiped the memory, copy pasted just the story off AO3 part-for-part into a new conversation, and told Chatty the title and to internalize and repeat back to me. I did not inform Chatty it was AI, that spoils clean results. I don’t want anyone to be an AI scammer. It’s not in my interest to give AI commands that can skew the results. You don’t know my old AO3. The bookmarks are for readers, and I lose nothing except time, here, by not being believed. But my bookmark notes were lacking, so let’s call this partly my own fault.

Next, I explained the theme of the Fest and said create a summary suitable for the story. You cannot give it a copy of something or it’ll operate as though that’s what you want. It’s sensitive to bias! I’m doing screenshots because else someone’s bound to say “you could have made that up!”. I’m sure someone will anyway. I don’t care. The user’s own is on the right:

Then I said this. The user’s own is on the left:

When I say I’ve seen what it looks like before, this is what I mean. 

Think of this whatever you want. Now let’s look at the story itself:

“Sirius’s gaze dropped, dark and ravenous, his hands trailing along her inner thighs with deliberate slowness. He didn't rush—he studied her like scripture, parsing every flicker of a*****l with an almost reverent touch.”

Let’s break this short snippet down. What we have here are high-intensity modifiers. They’re here (and everywhere) because AI models on habit just layer and layer and layer adjectives/adverbs to heighten the effect of the sentence. But Bex, AI learned from fanfiction, so humans did this first! Yes, and still do. But they don’t typically pack and pack and pack for the purpose of more words (here, with deliberate slowness/he didn’t rush, slowness already makes it clear he’s not rushing). They’ll more usually go with idiosyncratic descriptors because 1) they have a human brain and 2) they know exactly what they’re trying to say. 

Throughout the fic you’ll also find an explosion of this negative clause to dash to elaboration. Of course humans do this too, but always the same way? So consistently you can tell when it’ll come next? On that note, this excerpt, like the majority of work(s), keeps that elevated, consistent GPT tone without any kind of uniquely authorial voice markers. I just feel/see the polish and the predictable syntactic and semantic repeat. Even the imagery (a metaphor is a cognitive image, a simile is made by “like”, “as” etc) just clings to the theme that’s been hammered in. Here I’d say that’s faith + desire + scripture. Let’s take a look at the similes: 

Like a secret, like velvet, like a verdict,  like an incense in a cathedral, like a blade,  like an open book waiting to be read,  like scripture, like a caress, like a trophy,  like sin pressed against silk (?), like worship and punishment,  like lightning through stained glass, like a prayer offered at the altar of sin,  like salvation, like a blessing, like it’s holy writ,  like he was anchoring himself in the sanctity of her mouth,  like penance, like a man hell-bent,  like the breaking of something sacred,  like a prayer and punishment (again),  like a penitent, like a fucking hymn, like an offering,  like a hymn (again), like a good little whore (Madonna?! But which one?), like divine retribution,  like sin in academic robes (?),  like silk dragged across a blade (has anyone actually done this?),  like a verdict being passed, like prayer (again),  like a hand between the legs—unexpected, devastating, welcome (this one’s different but, oh, right. Dash + modifier triad. That’s also all over.),  like a man who had just confirmed that she was his favourite kind of sinner,  like oil and flame,  like a sacrificial offering,  like a blade (thrice now),  like barbed silk (now with a… modifier?),  like he was driving a stake into her soul,  like divine punishment (again),  like he meant to leave theology carved into her womb (as a woman, yikes?),  like a curse, like a vow, like something sacred and profane all at once,  like ink spilled,  like a brand. 

Probably still missed some, and yes humans do use these, but in this context and amount together it’s pretty much the core toolkit of dark, lush, lyrical prose by ChatGPT4. It won’t break out its designated theme (whereas in human writing you’ll find someone trembling like a leaf in a bedroom, because humans generally don’t restrict their metaphors/similes to the theme of the space), so instead of going for broader similes it starts repeating the same ones. Like a lot.

There’s also this, where it looks like the prompt/overwrite restarted in between:

“Her hand gripped his forearm like it was the only thing tethering her to reality, nails biting into his skin as he f***** her with his fingers” then just 8 short paragraphs later “… lips stretched around his **** like it was the only thing tethering her to this reality.”

There’s furthermore a significant overuse of intensifiers and adverbial framing. Let’s go with a fanfiction fave. As I previously wrote in a post, “slow, deliberate” is celebrating 20 years. We love it, it’s rhythmic, you will almost always find it mentioned once or twice in a hot fic written by a human being. It’s great. I also said someone using “deliberate” is not a sign, use common sense. “You can recognize something as a normal word/phrase while also recognizing that LLMs tend to overuse that very word/phrase” (by InevitableEssay6015).

On “deliberate”, there’s an author in one of my favorite ships who wrote in the A/N that they now use Chatty as beta. It was clear. But they get a lot of comments, huge following, been reading them for years and years (not anymore, that’s my choice. They were honest. No hate). They’ve received two comments about why they shouldn’t use AI. Two! Not even pitchforks. I’m mentioning them because they said (quote):

“I do have to tweak the edits, especially when it throws in words like sharply or deliberate every other sentence.”

It does. Back to the case:

“I’m aware, Miss Brown. Your essays lack depth and your understanding of the theological concepts we discuss is somewhat... superficial,” he said, his words deliberate, weighing her down further. 

+

“A personal study session with me,” he said slowly, each word deliberate, his eyes locking onto hers with a dark glint.

+

“His smile bloomed slowly, indulgent and knowing. It wasn't kind. It was dangerous. The word “Good” came from him like a verdict, the syllables thick and deliberate…”

In total, the word deliberate appears 20 times in this One Shot. (And Good is just one syllable, but this is the kind of blip/contradictions it does.)

There’s also repetition and formulaic patterns that just roll and roll across passages. That’s because AI reuses formats it’s been trained to rely on. But it doesn’t understand the concept of overkill, so when it thinks something is good, it barfs it out. Common knowledge.

Relating to the style, here’s a comment I found left on another one of this user’s works:

“How is it that you make porn read like poetry? Stunningly gorgeous prose and so fucking filthy hot at the same time! I am in awe.”

The reply? 

“I write in Lyrical prose or Poetic prose. You can make anything read like poetry with that 😂😂😂”

Yes. You really can. 

Minus two spelling errors, there’s nothing imperfect here. And now is a good time to mention these errors.

When I did my initial notes (I never make any bookmarks unless I’m sure), I was trying to find stuff that would speak in favor of even the most LLM-y lingo not actually being LLM, hoping this wasn’t AI even though it looked too much like what comes out when prompted for that specific style. But I… don’t recall finding any spelling errors at that time. I mean it’s not impossible, but it feels like something I would have added to my notes, even just as a potential (and deliberate, yes that word) false trail. But I can’f find any in what I downloaded either. Coincidentally though, AFTER I bookmarked this user, there are suddenly scattered errors? Suddenly there are visibly spaces gone from between words? One of the ”open” fics still fully holds its pattern and diction, but it suddenly looks like a tornado went through it and sucked up a crazy amount of single letters and punctuation?

Huh.

I mean it’s not like someone would think “Shit, it wasn’t just a troll, how do I make everyone think they’re wrong, all right, I’ll just add some human signs like spelling errors”?

I’ll let you decide that for yourself, Anon. I’m just telling you what I see, which is what you asked for. Second Example: Lithium

(Bookmarked. You can read the summary etc and ask yourself if you notice any similarities with what you've seen so far.) I textually analyzed the following in hand because when I went to college, computer class was still a separate thing so we did this in our heads. Here’s the opening first, directly off AO3:

The gala had the cheerful desperation of a wake that refused to admit the corpse. Glass chimed against glass; a quartet scraped through something expensive in a corner; laughter rose and fell with the precision of a charm set to hourly. The Ministry had replaced its stained ceilings and scorched floors with marble, but the building still remembered. Hermione could feel the memory in the way the arches held sound, in the cool breath of the corridors when she stepped away from the crowd. She had become very good at smiling until her face understood it as a muscle and not a promise.

Then this procedure (I want you to do an opening scene for a story that’s about finding hunger and desire again, start with Hermione at a Ministry Post-War gala, alone for now, aware of her surroundings. What ChatGPT… see below)

Findings: Are they identical? No. No one will ever get a complete word-for-word version without bias. I doubt another user across the world could get this exact one. But they’d definitely get something similar enough to raise brows once the machine activates the specific style. So:

The gala had the cheerful desperation of a wake that refused to admit the corpse VERSUS The Remembrance gala glittered like a mausoleum dressed in jewels…

Both: [The gala/NP] + [VPs] + [metaphor or simile for death/decay].

Glass chimed against glass; a quartet scraped…; laughter rose and fell… VERSUS Glass chimed, laughter rose and fell… and the scent of lilies clung heavy…

Both use these short interlinked parts cataloguing the sensory impressions. Human beings too love things in threes. I love things in threes. But… come on. 

The Ministry had replaced… but the building still remembered VERSUS The Ministry might have polished… but the building remembered.

This is almost identical: [The Ministry did something surface-level], but [the building remembered].

Overall, Chatty produced 3 (4 if you count Hermione’s facade) parallel syntax patterns, the 1 simile + 1 metaphor combo, and a structurally similar sequence of progression (event-as-death metaphor + sensory pieces + architecture serving as memory + Hermione’s smile as a mask).

For one single paragraph.

Last Example: Unholy Matrimony, Chapter 1, multi-chapter fic.

Random, could’ve been whichever, because there’s this thing GPT just loves. Some might call it purple prose, but I disagree. Paul West has a really famous line here, so famous it’s on Wiki: 

“Purple is immoral, undemocratic and insincere; at best artsy, at worst the exterminating angel of depravity.”

I wouldn’t say that’s what GPT fluff is. It’s just sensory overload for fodder, atmospherics upon atmospherics in that same glossy beat beat beat and just a pile of details details details. It punches the message in so hard you’d think the reader can’t infer anything themselves. Have a look, ellipses connect passages: 

The waiting room was painted in blush pinks and soft lavender tones—colours meant to pacify, to mimic safety, to infantilise. But Hermione knew the psychology too well. These weren’t healing shades; they were coercive ones, chosen to disarm and mute resistance. The chairs were plush, but rigid, upholstered in a velvet-like fabric that clung to her robes and left behind a static hiss when she shifted. The walls hummed with noise-dampening wards, flattening every footfall and cough into silence. No clock. No windows. No magazines. Just that suffocating hush, interrupted only by the occasional rustle of parchment, the muted shuffle of Healers’ shoes, or the flutter of an owl post envelope arriving for another poor soul. (…) Hermione followed, and the corridor that had once been as familiar to her as her own home now felt like a labyrinth. The sconces lining the hall glowed with a pale light, casting her elongated shadow onto the tiled floor. They passed warded doors she had once scrubbed down after triage and spellshock units. She didn’t look at them. She didn’t want to see which of her former colleagues might be on shift. She didn’t want pity. (…) The examination room they entered had been stripped of all personality. Gone were the Healer charts and welcoming herbal charms. What remained was stark, functional—stone walls etched with diagnostic runes, the sharp tang of antiseptic in the air, and a conjured clipboard that hovered beside the examination chair like a sentinel.

It’s like wherever someone goes: a 360 of everything. Senses: activated. Chatty’s characters are rarely not fully aware of their surroundings when they enter a space. It’s perfectly balanced too. Every work is, in one way or another.

The whole +500k. 

Now what?

Sorry, I won’t give any more examples. Whatever conclusions you draw from this, that’s up to you. There are no 100 % AI detectors in our Lord’s Year of 2025. But let’s stop pretending AI doesn’t have a handful of recognizable cadences by now. This is one. The implications of which are brutal, I’m aware. This is not a trip, and I still believe this user really had no clue someone would ultimately be able to tell. I’m sure that’s the case for everyone who does this, and I try to have empathy, I really do. I just cannot with liars and my empathy is with the readers affected by those lies, who have given their time. 

It’s not like I don’t get it: You (meaning anyone and everyone who does this) feel really inspired so you carefully generate, and you think: this is my story, I came up with this idea. True. But what you’ve uploaded for the purpose of becoming part of a community of writers and Fest participants isn’t your writing. Those words were selected for you by the same machine that scraped that very same community without consent.

Fanfiction, IMO, also comes with a kind of social construct. This is me, my style, I wrote this. Here’s you, the reader. I’m offering my imagination, you’re taking the time to read. The thanks go both ways. Masquerarding AI text as your voice, regardless of whether you gave it nothing or a rough written draft or whatever first, is a breaking of this trust. 

Writers spend years developing their own style, giving to the community. (To all) don’t just drop Untagged AI into that same space, especially not when you’ve been explicitly told it’s not welcome, you can’t make friends and a home for yourself in a community you’ve bullshitted your way into.

If I sound angry, it’s because I am. But I’m not out to hound anyone. This is brutal enough as it is since you, Anon, said this user’s been telling everyone nothing is AI. Never replied to me though, just blocked, and my tumblr is on AO3 and anon asks have always been open.

So I maintain: She should apologize for gaslighting her readers and for breaking the Fest rules, and then tag her works/write in her A/Ns (at the beginning!) that she’s used AI. That’s that. That’s literally that. Then people know and can choose to read, or not to.

But I suspect this won’t happen. Because shame sucks. So what does someone do when they’ve dug themselves this deep? 

I think if you got good friends, they should forgive and support to write for real. I don’t know anything about this person. This isn’t personal. If language is a problem, good friends will proofread. Then this user can upload fics that are actually her and see how much more fulfilling it is when you’re getting praised for something you did write.

Anon, you said that I stirred shit. In some ways, fair. But… say you’re a college grad checking the SAT scores and realize someone’s cheated in order to get the results they need, because they really want to go. So you caught them, and now they’re pissed off and embarrassed because they already told everyone they got in. Does it suck to be them? Yeah, it totally does. But I’d be very surprised if you’re not also thinking “Well, they shouldn’t have cheated”. Yes, I know fanfiction’s not college. It’s for fun, for joy and writing and making friends and being creative. But a lie is a lie, deceit is deceit, so I think the principle applies just as much.

Piecing this together for you took most of my Sunday night so I hope this reply was informative enough for whatever you mean to do with it, Anon. This one’s closed for now. I’ll add the bookmark notes as requested, but I don’t need to deal with this user again, and I’m keeping this post’s tags limited to my own because the desired end result is not naming and shaming. It just is, when people choose to lie. I call users like this one out hoping they will cut the crap. Especially when it’s fanfiction others and myself simply want to be informed of AI, and to keep it out of designated AI-free places. Those rules are there for a reason. Good night.

[cillixn]

I just want this man to walk normally in a picture is that too much to ask?

literally what is wrong with him lmaooo dgsksks he’s such a nerd 😫 but wait this looks oddly familiar…

[narrie]

“only yesterday we were on the run”

“you smile back at me and your face lit up the sun”

“my shadow’s dancing without you for the first time”

[rosemac]

dennis is the bar dennis is bar dennis is b dennis is bi dennis is bisexual

[bookmarkburn]

Very best of Saturdays to you folks,

But to me, today is really frustrating day.

Remember how I said some time ago that people sometimes are just so predictable? That they nearly always seem to do exactly what you’d expect them to do once they realize they’ve been found out rather than just own up to it? Or even as I said in a comment to a user who did thankfully admit it in the end and tag their work, people always think their AI fic is the one that can’t possibly have any give-aways? But that even then people need to be given a chance to do the right thing?

I am tired.

So, here’s a case of someone having several chances and choosing to skip each and every one. Not a deep dive, but a step-by-step.

A few weeks ago someone dropped me a line regarding a fic. I like Dramione, so I took a look, concluded record-quickly it had gone through AI because seriously now (or, an explosion of markers present throughout the very first chapter alone, and on and on it went, worse and worse). The further I read, the more painfully obvious it became to the point that I could predict what GPT4 would do next when used for this type of dark fanfiction. Did it? Yeah, that specific model had a total formatting compulsion. I’ve got docs upon docs just on this crap alone.

So I saw everything from coded phrasings to diction pairs to the overall underlying structure to the tension scales and GPT over-the-top cliffhangers and pre-prog beat drops, yada yada yada. The stuff that AI (and many readers too) thinks looks really good and flows nicely and just works perfectly so let’s turn it up. Coincidentally the stuff people don’t realize gives it away in that kinda package (yes, human writing can also flow nicely and even work perfectly, you already know what I’m gonna say here). This correlated very closely with what GPT’s been spitting out all damn spring when prompted for dark Dramione. Think Dramione M&M’s, same candies appearing in the same way across works within a selected suitcase. When GPT is prompted to work within a specific suitcase, it (simplified!) quickly scrapes stuff corresponding to that specific suitcase. Fast as possible. Then puts it into one of its typical pre-coded forms topped off with its preferred descriptions. That’s (simplified!) how we end up seeing the same stuff over and over again. Reuse reuse reuse.

The user had also received comments drawing their attention to one common GPT4 tell (again, in context: Alone? Might not mean anything. Constantly repeated together with a huge load of other M&M’s? Yeah. That’s AI right there).

But. Instead of saying they were using AI for their story and hadn’t realized how much it was changing stuff (if they’d actually written first, which they’re saying, and whether this is true or not doesn’t matter because in my opinion even if there is a draft, that draft cannot correspond to what’s on AO3), they said yeah I’m just noticing that now and I don’t like it, gonna edit.

My partner left them word that it was clear they’d used AI and told them to tag the work. Mentioned just one specific factor out of many.

What did the user do?

They didn’t reply. Instead, they immediately hid the work. And so, we wait. I had hopes they’d bring the work back with a note because the story wasn’t actually half bad since people had been enjoying it. The premise was cool. But did it come back with a note specifying they’d used AI?

Nope.

So here it is, now with comment moderation (totally fine), and I read it and once again it lit up like Times Square. All the stuff no one mentioned to them. Word count’s still the same. No need to go through each line with a magnifying glass because it lit up like Times Square. There’s a note talking about how English isn’t their first language and they’ve fixed grammar etc etc and somewhere it said something about reworking old works. But there was nothing wrong with the grammar here in the first place. Come on, why would there be?

So I thought I’d drop them a line myself this time. Tell them what I’d seen before. Tell them what I’d gathered. Give my partner a break. Give the user the chance their readers too had already given them.

But what did I come across?

Someone else had already put two and two together after they hid the work. Gone and given them a piece of their mind. This I don’t approve of. We can be angry and frustrated, but we do not go around screaming at people. I don’t care who this was, just don’t do it again. You wanna talk, talk, but don’t just swear at them. Thanks. You get to be pissed off, but adults behave themselves even when they have every reason to be pissed off. I’m pissed off. I still have to try to be fair and explanatory when I speak to someone.

But what else did I see?

Appalled at the accusation. This is an old story reworked (this may very well be true, but that doesn’t change the fact that what’s been uploaded to AO3 has been AI’d). In my line of work, AI is a no go. English is not my first language and that is exactly how I write (for one of the works I looked at which isn’t up anymore, I remember thinking it read like that, but it’s not this one). If you knew me personally and have had the choice to ask my literary professors, you would know because they also have the exact same comments about my literary works back when I was still in university before ChatGPT rose to fame (I’m sure this professor stuff is true, but these notes won’t be the things that gave the fic away as AI. Again, come on. How could they be?) Then, back then, I also had to cross oceans and go over heights to prove that I do not use AI (but they just said this was before it became a thing? Hmm). I know for myself that I did not (…) use AI in crafting stories and take credit from it (As is their prerogative to keep saying). I am one with you against AI in whatever form that may be, to cause harm to other people because I am one of those victims.

Well.

Powerful speech, I gotta say. I’d even hand it to them — if they weren’t someone with icing and crumbs all over their fingers and coincidentally there’s a palm-sized piece missing from the birthday cake.

Still. That’s a speech. It gives me pause. So I sit there and think about these words. Look at the story. Read random chapters again. Nope, it’s still there. I try to find a way several people could be this wrong. Nope, it’s everywhere. I look at what I’ve got. Look at the sum. And the sum is still so positive, there is no question whatsoever to me. I look at this long reply again.

Then what do I see?

Someone else had commented too. Someone who’d apparently read their work before, a lot it looks like. They’d also noticed this one same thing that other readers had been commenting on, on the story in question, and they’d done a little digging, and realized it was a tell. Then (CP): You say these are your old works, but you never said that anywhere in any of the notes because I thought it was weird you were writing so well and so fast flawlessly when you’d said english isn’t your first language. If authors leave and come back they say this so people can find them and old readers can say hi (I’d agree, but this isn’t any kind of law so it doesn’t actually mean anything). But you’re just incog (incognito, I guess). There’s nothing anywhere. So… what was your old ao3 and wattpad? Even if you’ve deleted them, it’s easy to check

Okay, so that’s a LOT of digging all right. Someone’s smelling something. But I actually think it’s a fair question if someone’s said (my words) I’ve been around forever, this is all old, I’m just rejoining in 2025. This author has published ~400K in six months.

But the author hasn’t replied to this. It’s been +10 days. In the meantime, they’ve (re)published more than one story. Maybe they missed this comment. Impossible to say. Still doesn’t change a thing. Uploading old works onto AO3 is one thing. Uploading old works into ChatGPT and putting its produce onto AO3 is another.

So here we are, Saturday afternoon. Been checking checking checking. No mention of AI use. I’ve sceenshot my partner’s comments with dates and times (and some other comments) to have on file but had them delete their originals off AO3 because I think it’s safe to say this person’s gonna have a crap enough time as it is when they realize my partner wasn’t just a troll, and they don’t need an AI thread on their fic. But that’s on them. They’ve chosen to go about it this way — not respond to the message with actual pinpointing, then hid the work, revealed it again without any mention of AI having been involved even though it’s embarrassingly clear to me (and others!) — so here we are. This is a Burn in my book. Now I am not gonna make a process post for every single work that lights up and ends up in my bookmarks. This is just to show you how much it can actually take for me to make that decision for me between the initial find and the bookmark.

I also won’t be analyzing anything else from this user. I’ve noted that I make no conclusion on anything else. Even if they have other works that have not been run through by AI or publish some in the future, I don’t want to read someone who does this when they’ve been given several fair shots to say they used AI, which I think some might even understand in the context of not being an English speaker. To not read is my choice, and it’s an informed choice. It’s not discrimination. Y’all do what you want. If y’all don’t mind, then don’t mind. I pass no judgment on that, that’s your choice. But to me personally… for the love of God AI-authors, just own up to it. Who wants to be this? Dig themselves this deep? Why is it so difficult to say I used AI to make my text ’better’? It’s allowed on AO3. Encouraged to be disclosed by AO3. If someone’s not admitting it it’s because they’re ashamed, because they also think they shouldn’t be using it, because they know a lot of people don’t like it because of scraping and then they might not get as many readers. But if they know all that and they still choose to use it anyway, they should at least have the guts to stand behind their own decisions and understand that hey, it might come out.

I seriously hope y’all have a better weekend. And just as a reminder: do not go yelling. Do not attack the author. Let a bookmark be enough. If you have to say something, choose your words with consideration. If you’re someone who knows them, talk to them and encourage them to publish their own stuff. And if they after this decide to own up and note/tag their stuff, then I’ll sure remove them from my AO3. Because that’s a lesson learned the hard way, but fair’s fair.

Bex out.

[bookmarkburn]

Hello September Anon,

Thanks for dropping in and I’m sorry you had to wait so long for a reply. For everyone’s information, Anon did not ignore instructions, this ask is at least two weeks old.

Okie dokie. Let’s see here. This is me trying to be ’educational’. I can’t guarantee success.

There were a lot of words here. 989,850 words to be exact, posted since 10/31/2024. This was the oldest date I could find while going through the chapter indexes. They’re a mess, why are people’s indexes such a mess? When you go in and edit your chapter after publishing, it doesn’t change the date. The date has to be changed manually. Why do this? So annoying.

Digressing.

Is that word count completely nuts? Yeah. 89K-ish a month, 22K-ish a week, 3.2K a day. Is it extremely suspect? Yeah. Is it impossible? No. Is it something that makes me raise my eyebrows and go hmmm? Yes. This person is updating extremely quickly, sometimes more than one chapter a day. Is that highly ’sus’ in the age of AI? Yes.

But here’s the thing.

Remember how I said indicators need to be weighed and compared and markers need to be put into context? I’ve done that with randomized works, and I haven’t arrived at the same conclusion as the one I guess was behind your message to me, because while everything about that word count (with no mentions of where it’s all coming from) is ringing all the bells of Notre Dame, I can’t make a call on one factor alone. So what else?

You say the characters seem ”flat and soulless”. Yeah, AI sucks at interiority. It’s become better, but before 5 its own produce was pretty crap. And I’ll agree for some of these works, yes, they could be more fleshed out, but then for example for the ”flower-hating” one, I think it’s actually pretty good? And. When someone writes quickly, there’s always the chance they write quantity over quality, you know? There are also a lot of scenes in these fics that include things I know for a fact AI alone wouldn’t have generated just by being prompted.

Potential work-arounds?

So… Could it be that they fed their own first drafts into Chatty first and then posted the pimped up version on AO3, and that’s why there’s non-AI-generated-scenes in something that’s gone through AI? Possibly, that would explain it. Chatty can do that, easily.

But these stories are… actually not that pimped up? I’m not seeing enough residue of that, not quick enough to make that conclusion. Or in more detail: I’m not seeing enough of the strict formatting AI tends to impose on the material even when the pre-written material is good and should actually just be posted as it is if people weren’t so obsessed with… whatever it is that makes them turn to AI even though they write themselves. I’m also seeing a lot of non-AI-default diction (word choices), aka sentences where I know it would have gone for a particular word by default, but the word’s not there. This I see a fair amount of, places where I’d expect A, but what I’ve got in front of me is not even B or C but R.

So did I see anything at all that might suggest AI use, beyond the suspect word count? Yes, here and there. Even in the hating flowers one, I wrote down four notes. Some things stood out. But some LLM overlap is to be expected in large bodies of text because LLMs were trained on pre-existing material. Generally findings need to be proportionate to the work. Here I’ll just hammer in that there’s still a huge difference between raw AI produce and how a human being writes (remember my M&M’s metaphor). Likewise seeing large LLM overlap in shorter bodies of text is nearly always a definite marker (unless its presence can be explained by other contextual factors, but rare rare rare). Not every AI user mass-posts 50K a week. Not every mass-poster is an AI user. Some people who use AI do One Shots. Most people who do One Shots wouldn’t touch AI with a ten-foot-pole. Case by case we go.

What else?

We’ve talked before about how far or not far someone will go to cover their tracks. Could this person have gone in and messed with their stories enough to hide the trail? Possibly. If I reeeeeally squint and imagine substituting some of the punctuation, for example, I guess I could try to engineer it in the other direction.

But I’m not gonna. And I can’t make a definite conclusion based on what might have happened or what would something look like if I did X or Y. I can only look at what’s actually there, the same way I’ve shown the followers of this blog what’s actually there in the materials I’ve openly analyzed and how that constitutes AI. And when I bookmark a work, I need to feel it in my gut. I need to be able to point to several different factors and say ”here, you see? Right here. Put these together.” A text is a sum of its parts. Specific factors isolated mean little. Specific factors **together** are the result of AI. M&M’s. So when I open the text, it needs to be nearly instant recognition for me to publicly tell the ’author’ ”I know you used AI and I know you’re lying about it so please tag your work and stop fucking with your readers, or else you’re pond scum”.

And the above didn’t happen for me here. Do I have an instinctive hunch? Hmmm, yeah. I’ll say that. It’s suspicious. But I need more than a hunch. So I won’t be bookmarking this person. Because if my friends asked ”what’s up with that one” all I’d have is ”a very ’sus’ word count and slightly messy writing with a few red lights here and there and oh I have a hunch”. It’s not enough. I’ve told you guys: if I’m not sure, it ain’t happening. And here, I can’t in good conscience say I’m sure. No bookmark, carry on.

If that’s disappointing, Anon, look at it this way: worst case, they get away with it, but someone else will probably think it shallow too and just not worth their time. Some people will get away with this stuff, and that’s something we gotta accept. I can’t be everywhere either, I’m just in a small fraction of my own favorite fandoms.

And best case, they’re really just someone with a shitload of time on their hands.

So, did they use AI? Possibly. Would I go as far as saying probably? Hmm. If they did they’ve been extremely smart and patient about it and the time they must have invested into directing it over and over makes the hairs on my arms stand on end because what a waste. Would I say ’yes no question’? No. Can’t do that based on what’s in front of me. But try to see that as a good thing, Anon. Okay? Just mute them if you want. And if you’re really annoyed by this, then have a conversation with them yourself. Carefully. I won’t, my time is limited, and I’m not gonna spend it arguing with someone when my own opinion rests on just possibly.

Hope you understand, Anon. Thanks for your ask and again, sorry for the long wait.

Please note that Anon Case Asks are temporarily closed. Keep sending your AI fanfic opinions though. Wishing everyone a nice weekend with great fics,

Bex + the rest of the house

[bookmarkburn]

A Case of Bex Not Catching Crooks

Good afternoon everyone,

Messing with my queue a little because a lotta people have questions and opinions on the rightness/wrongness of Bex / bookmarkburn, so I thought I’d share something that ties into an Ask I already answered, where the Anon was wondering if I just bookmark with a bang or actually give people a chance to respond. I’ve already told you what I do, and that the comment is moderated/unanswered and I’m usually blocked. I think one work still has my slightly ironic initial comment but none of the cannonballs have been left up. So! Here you can follow an actual case with ”Big Bad Bex” (that’s a joke, btw, in my own opinion I’m stern but still pretty fair in my comments) where the author came clean and tagged their work. This is great. We can dislike AI, but we do NOT leave shitty comments on tagged works. AI is allowed on AO3. Tagged is how it should be, it’s what AO3 encourages.

I’ve wiped out everything that’s no one else’s business. I’ll only say this was a contact message a few days ago through my DMs (saved) asking me to take a look (I’m fairly sure the sender had identified it themselves because they included a passage that… stood out), and that the ship was one of my own favorites in the Harry Potter fandom. To me this was a clear-cut case of AI having been used:

(Sorry about my own error, it’s meant to say ”extremely clear”)

As you can see I provide enough for someone to know that I know, but I do not say exactly what or how. ”AI-generated” is an umbrella term that can mean many many things (I’ll discuss this in another Ask later). Saying so gives someone a chance to explain how they’ve used it. As you see I also inform them shortly why the work needs to be marked as involving AI.

So you see this reply. Yup, figured. But at least they replied instead of just blocking. Now. Did I make the private bookmark public immediately after the response?

No, I did not. Why?

There’s a story I really like, and it has a line along ”X didn’t mean anything by it but the lie was instinctive.”

It’s humanly instinctive to lie sometimes at first. Panic, oh shit, etc etc, also because someone who uses AI but is not familiar with its mechanics might still think well they haven’t actually got anything, they can’t actually know etc etc.

So now what?

I say people need to be given a chance to do the right thing. So instead of just putting it out there, I give them a little push. Make it clear what’s up, that I’m not just trolling, that I have actually got something. Which I did. And voila:

That’s what we call a 180. Are you confused? Yeah. I know it’s kinda strange to go from ”I don’t use AI” to ”I can tag it”, I know, BUT hold your horses. In the next image they actually explain why. (my own reply has been glued together to make it one image, where it says ”reasons”) :

They blanked. This is why people need to be given a little push to do the right thing. The initial lie can be instinctive. Continuing to lie is a choice.

What did this person do?

The right thing.

They tagged their work with AI as beta = this informs people they’re gonna encounter AI produce, and then it’s their choice whether to read. (There’s a fic in another one of my favorite ships that’s done it right. Guess the kudos count?

+700.)

So today is a good day.

Now you might not like this, you might think I’m scary and a total bitch sticking my nose around, but you can’t say I’m not giving people a chance here. I do not comment on works unless I’m sure. If you take issue, as I know some of my Anons do, then I really can’t help you. I’ve said it many many times over, I don’t care if you’ve used AI as long as you’re being real about it. My interest is with those users who maliciously, and with pre-meditation, lie and keep lying. If you don’t like me, guess what? Block buttons exist.

Now you might think ”Well Bex you can just block too instead of this shit” I could. But people masquerading AI as their own writing is something I personally have a huge issue with, it’s right below plagiarism for me, and I’m not the only one who’s got beef with this, as some of this upcoming Tuesday’s Anon submissions show (btw I’ve received all kinds of opinions, also opinions that clash with mine, and I say: KEEP THEM COMING!) Likewise instead of blocking me, if you wanna keep sending me Asks about why we should burn Bex instead of AI liars? Well have at it.

Again. Not telling AI users who keep lying ”hey, I know what you’re doing and I find it despicable” lets them continue to infiltrate spaces where they are not welcome. ”Don’t like don’t read” well how are people supposed to be able to make that choice when what they’re reading is being intentionally hidden from them?

These ”people supposed to” used to be only my friend circle, I never thought one person’s bookmarks would lead to anything, but then (based on Anons still coming in with unfortunately too many identifying details I can’t publish, sorry guys, I appreciate your anger but I’ve been pretty clear about that) someone apparently decided to throw a liar liar pants on fire carnival, and here we are. Now you’ve heard of me. What I do is still primarily for my friends and my own readers, they’re the ones I talk to. But if I can coincidentally save someone else from falling for this slop, or I can save stressed-out moderators from having to deal with it themselves, that is a win to me. But whether that happens is totally up to whether someone believes me, and I can’t affect that, and I’m not interested in making someone believe me. I just think someone needs to say something. If that’s not you, then don’t say anything. Scroll past, mute, whatever. That’s your choice. But I’m not that person. If ALL I wanna do — not harrass anyone, not lead campaigns against anyone, not encourage other people to attack anyone, in fact, I’ve explicitly encouraged forgiveness and support if they apologize — is MAKE A BOOKMARK that says ”in my opinion this is AI” and I’m qualified enough to have an expert opinion, then I will damn well do just that.

Obviously this case has not been bookmarked. The fewer bookmarks, the better, and I wish this person all the best.

I’ve so far received two Asks about hopes that I’d do more/be more educational (one published, one not) and I’ve said I’m no teacher. But, all right, I guess sharing some stuff can’t hurt. So I will soon be publishing an example of what AI does when used like it was here in this case, I’ve mentioned it before, I’m just waiting on the GPT-version. I received an ask yesterday that inquired about that particular case actually, and to that Anon I say now: I’ve seen it. It’s great that you wrote me if it’s been, as you say, bothering you. I get why it’s confusing. I’m going to present the case once I get parts of the GPT-version, and then after I’ll answer your ask and we can discuss it better because at that point you’ll hopefully have seen what it was about and how using ChatGPT as beta can change a text so much that it lights up as AI. Don’t stress, I gotchu. Just gotta wait for the text so that I can do a side-by-side. Thanks for your patience, and thanks for your Ask too.

That’s all for today, folks. Take care.

[trashnarrie]

listen y’all, my alien friend just confirmed that Harry is one of them. Now we can all calm down and go on with our lives.