Bex / Bookmarkburn

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.

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,

(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.

Any length. Any opinion will be treated respectfully. We’re aiming for conversation here, folks. Not bashing. So drop me an (Anon) ask starting with ”AI fanfic” or send me a DM starting with the same line, and say as much or as little as you want. I’ll make a long post with the answers on Tuesday 10/7/2025. DMs sent will never have usernames revealed.

Remember asks and DMs are open for general thoughts or rants or opinions or whatever. Some people have already sent some that require no input from me. I’ve got two queued. I don’t know if inviting conversation is the kind of ”change” my today’s polite and thoughtful Anon had in mind (and I’m not done thinking about your ask), but I’ll start here and we’ll see how it goes.

I personally find it hard to believe that so many people read these works and don't realize they're AI - it's pretty fucking obvious 

I think it’s obvious if you know what to look for. Sometimes ignorance is bliss and ignorance rarely has any intent except for existence. Many readers aren’t native speakers, and that makes it even harder. 

And of course, not all AI works will be bad storywise. People can get pulled in, and when you’re immersed, you don’t necessarily pay too much attention to how something is presented, only to what is presented. 

I’m currently sitting on a story that’s been very popular and received a huge amount of praising comments. No one seems to be questioning the ridiculously quick updates, the repetitive sentence structures appearing in nearly the exact same place in each chapter etc etc. And I can’t blame them. But it also doesn’t feel good to expose, so I’m gonna sit on it a little longer before contacting the author. It would just be so much easier if people were honest.

So, I think if you're going to make such bold accusations that could ruin a person's reputation, you should at least provide examples, evidence, an analysis maybe?

Are they? Bold accusations that ruin a person’s reputation? Let’s get one thing straight. People who choose to use generative AI to produce fanfiction and post it online as their own writing ruin their own reputations because they think they can outsmart everyone else. As for examples, evidence, analysis: condensed examples of findings have been added, as requested, here. See partial analysis of one bookmarked user here. This analysis includes enough of what I personally consider evidence (there were mountains more) because I’ve said this at least twice now: there are no AI detector’s in our Lord’s year of 2025 that will give a definite result and I don’t like those anyway. I do manual analysis, for which I’ve been trained. It’s your own choice whether you believe me or not, I don’t bookmark anyone unless I’m sure and would stand up in a hypothetical court of law and declare ’your honor, it’s my professional opinion that this text has been AI-generated/AI-processed, because of these and these…’ (I don’t see this happening. Plagiarism is a thing though.)

Because I'll be honest with you, I see what you're talking about on a few of the works you've bookmarked. It's the excessive use of adverbs and adjectives, the metaphors that don't mean anything, the repeated sentence structures, blah blah blah, I do see it. 

That’s really great. Just remember that each and every one of these isolated doesn’t automatically mean a text is AI. Each text is a sum of its parts. Let’s not everybody lose our shit because someone likes modifiers ! Findings need to be put into context and compared as a whole to known AI produce. Permission to utilize common sense is still in effect.

Would I call it out? No. Because of two things: 1 when it's that obvious, I think the trash will take itself out. No one can bear to read that kind of prose for long, it's shit, and none of those fics will do well and earn much attention because it's so obviously not human it's uncomfortable to read and readers are not stupid, they catch on to that. 

Yeah. You’d think, wouldn’t you. You know what I think? I think we can conclude by now that the ’trash’ (to use your word) does in fact not take itself out. Not even when it’s shown where the door is. As for prose being shit, let’s just pause for a second here.

Not everything that AI generates is seemingly trash, especially not when it’s got a smart human directing it. There is currently a multi-chapter fic in my private bookmarks that’s been extremely popular. The comments are raving about the prose etc etc etc. Nothing in these comments suggests that someone’s realized what it is (although the user might have deleted the comments). And believe it or not, it actually sucks to have to have that conversation, but I’m gonna. Contrary to what you might think, Anon, my thoughts when I trip on an AI-generated fic aren’t ’I got you now, motherfucker’. They’re ’why, why, why, why the hell did you have to do this? Why couldn’t you just write yourself?’ 

And 2 I think calling people out publicly gives yet another opportunity for AI to ruin fandom spaces. It's already harmful, but I think responding to it that way makes it even MORE harmful because it creates an unstable and unsafe environment where people are afraid of sharing their writing because they think they're going to get called out, even if they didn't do anything. I see authors afraid of this everytime this gets brought up on reddit and discord, and even AO3 seems to agree with this, as this is one of the reasons why they decided not to ban AI on the website.

(Among other reasons AO3 hasn’t banned AI, because they don’t have the resources for all the fics that would get reported and because again no detection software’s 100%)

I disagree on the first half of this. There’s actually an ’opinion’ ask in my inbox that addresses this and I’ve queued it because it’s about a lof of the same stuff I’m telling you here. As for these authors you say are afraid: What exactly are they afraid of? That someone is gonna comment ’this is AI’ because they write well? And when asked why they think that, this someone’s gonna reply ’well you always use the m-dash’ and the author’s whole world just… crumbles over ’—’ ? Somebody call the cops, we need to charge Emily Dickinson for using AI!

I’m sorry. I don’t mean to be condescending. Being scared is no fun. But… my dear brother or sister or however-you-identify in Christ: Writing well and writing like an LLM are two different things. People who don’t use AI to write got nothing to flip out about. If they’re flipping out every two seconds, you should ask them why. ’I’m scared bookmarkburn’s gonna get me’ well, I think we can all agree that bookmarkburn doesn’t come for you unless you’re using AI. Bookmarkburn might actually leave your non-AI work a really nice comment, she’s known to do that too. So chill out, please. 

But fine, if you want to call people out, I'm not going to tell you what to do. Honestly, AI is a new thing and we do not yet know how to deal with it and I can accept the possibility that I'm wrong and that we should be calling people out, maybe.

Mm-hm.

My problem with the way you're doing it still stands because you do nothing but point to a work and say: "Hey look! This is AI!" And not all of your pointed fingers are to obvious AI slop. I read all of the works you bookmarked, and while I agree with your judgment on some of those, there are one or two there that I don't think are AI. I think there's very obvious evidence of human writing there.

I’m gonna disregard this first part because I’m the one who hasn’t had time to answer your ask previously, and I’m not gonna chew you out for commenting on stuff that’s already been fixed when it wasn’t fixed at the time you submitted this.

I’ve already said that I’m fully aware of the issues with the initial bookmarks. I apologize for not immediately adding details, I seriously didn’t think a single person’s bookmarks would lead to me having to create an actual blog. But I will not apologize for how angry I sound. I am angry. You should be too. Actually, why aren’t you?

As for the second part: I’m gonna take a deep breath here.

All right.

People who use AI to create fanfic are not all some pimply-faced teenagers sitting in their mom’s basement with one hand down their pants and an IQ of 65. They’re everything from college students to housewives to bored mansion-owners in Boca. A fanfic being AI does not mean a human being hasn’t touched anything about that fic. In fact, it’s extremely rare to see unedited ChatGPT output the way we saw here.

AI users are not all idiots. Thinking no one will catch on and that they’re outsmarting everyone else might make them seem like an idiot in hindsight, but many, many of these people are clever. They’re just also lazy and horny for kudos, but having a story in your head and feeling like you’re gonna explode if you don’t get it out is NOT an excuse to present it as your own work to your audience. And when comments are allowed and a user gaslights readers in the comments and advertises when the next chapter will be up, THAT is someone looking for an audience for their slop.

Where their cleverness falls short (for some, unfortunately not for all all) is surface level edits that do trick the immersed trusting reader. What most of these ’authors’ don’t understand is that there’s an underlying system and tweaking a word or two, or even having a beta reader change a little something, doesn’t actually remove the building-blocks of the text. There’s an ask in my inbox wondering what I mean when I say prompted/generated etc etc, aka different types of AI texts, so I’ll address that when I’ve got time. There’s also an ask about a user who’s been smart enough to foresee this and guess what? I won’t be able to burn them. I’m not totally done analyzing yet, but right now it looks like they’ll be getting away with it.

Now, could I be wrong? Absolutely! I do have daily contact with AI (work in advertising, you know how corporate is nowadays, AI use is required to boost your productivity and I can't afford to argue on principle, I've got kids to feed) and I'm pretty confident I'm able to catch on the clues, most of the time. But I still could be wrong, because I am human, and I am fallible.

Yes, I do know. You gotta do what you gotta do. Not gonna comment on your personal life beyond that. 

So are you, though. And I get that you are confident enough on your skills (I don't, really. I could never, I'm always hyper aware of my shortcomings) but I think as a community, we're going to need more than that.

I’m fallible in many ways, I’m sure. This generally isn’t one of them though. I get why you would think that. But do you also tell the pilot when you board a plane ’what if today’s the day the autopilot goes up in flames and then you lose control of the stick? You’re just human after all’? (No need to answer. I know the answer is ’no’, and you know why I asked.)

It also appears as though you at the same time think I expect people to automatically take my word as some kind of gospel. If that were the case then all this would look a whole lot different. Do you see any slandering campaigns? Am I loading up my RAM and going to war? Have I gone after the readers leaving love on these AI fics?

No.

You are free to keep reading bookmarked users. They’re free to keep using AI. The only difference is now they know they’ve been found out. I hope this will encourage them to either stop using AI or to tag their work as they should have from the very start. There are hundreds of AI-tags to choose from depending on the extent of the use. There’s no excuse here to leave it out, except masquerading as skill. But I suspect they won’t because that would mean having to tell their friends and readers ”okay, so, I lied to you. I did use it. Sorry, I shouldn’t have, I just didn’t think anyone would know”. So they get to chill in my bookmarks, and the only reason they’re there is themselves. If they want me to remove them, they know where I am.

I believe that you are a well intentioned person who really cares about creating a safe fandom space, and I want to say that so am I. I'm questioning these things because I care about fandom and I don't want this to turn into a witch hunt with people being accused without basis because of one single person's opinion about their writing. I believe this is harmful and it scares me and I think we need to do better than that.

You seem to be operating under the assumption that this is all just one single pissed-off person who’s just prowling the feeds of AO3 looking for her next victim. That might be my fault. But I’m not. I didn’t create a separate account called bookmarkburn with the intention of doing this. The account was actually called gr33nlights, and I created it for me and my partner this winter to share fics we’re reading. There are comments upon positive comments across AO3 from both of us. My partner’s not American, so you can actually tell if it’s me or them commenting. The username became bookmarkburn this summer when I got fucking tired of not being able to scroll my ships without seeing untagged slop left and right when I already have to deal with that crap at work and in my free time I wanna be with my partner and my friends and take care of our animals and write and read some goddamn fanfic. Not AI-fic.

I’ve also been in fandom for many, many years. I’ve published +200K. This means I have friends, loyal readers etc. Many of these know what I do. Many of these are smart. Many of these know they can text me and say ’hey, would you and [partner’s name] check this out’. I will not be disclosing my partner's name or username and this is not up for discussion. 

So: without basis because of one single person’s opinion is really not it. 

So please, if you could clarify for us: what are your methods? What do you see in each of those works that make you believe so surely they were created by AI? If you could provide examples taken out of each one, that would be great.

Already done. Y’all are welcome to ask for more, which I will then address as a public post. I promise to be just as thorough as I was here, but remember this takes time. I know, Anon, that you’re not in my head where it’s all clear and that’s not your fault, so I’m gonna try something here. And this is extremely simplified, okay? I’m sure Open AI would electric-chair me on the spot, but I’m just trying to do an easy digestible example for you. 

Do you like M&M’s? I love ’em. Human prose is like M&M’s with a zillion different colors. Authors share some M&M’s with each other. What an LLM has is like its own bag of ’best M&M’s’ that it’s put together from millions and millions and millions of texts. Millions and millions and millions of M&M’s. It’s been trained ’these M&M’s (specific syntactic structures, diction, narrative arcs, styles, etc etc etc, long-ass list, including shit it consistently fucks up) are the best ones to use, but because it can’t (yet) understand when it goes overboard, it compulsively just barfs these same M&M’s out. Different colors in abundance depending on what kind of text it’s meant to create (advertising? journal article? cv? essay?). So in the same kinds of texts, you’ll find these same-colored M&M’s consistently throughout AI-generated slop. 

Now picture this: if human prose is made up of a zillion different colored M&M’s, how are the exact same M&M’s suddenly appearing across a multitude of authors’ works in our year of 2025? In exactly the same way within style-sets?

Sure, there are trends. But just think about it. Again, way simplified. But think about it.

Also, when you call those authors out, do you give them the chance of trying to counter your argument or defend their process? 

Y’all have the choice to respond. Block appears to be the popular choice no later than after my second contact, usually after the first.

If an author were to provide evidence that their work wasn't generated on AI (say for example the saved history and alterations made on the document they worked on) would that be enough for you to consider you might be wrong on your judgment? 

Again, I consider this void because this is seriously just… not a problem? Using AI doesn’t mean there’s no document, it can be filled section-by-section through copypaste. Using AI doesn’t mean the work hasn’t had a proofreader, it just means the proofreader wasn’t able to see what they were actually looking at. Not their fault either, we wanna trust people. Take this:

I’ve just spent half the week on an older case where it was made clear that the story had been drafted part-for-part. Like any human author. Guess what? The AO3 version was ChatGPT-processed! Not just my assessment. The author’s admission. I’d commented a while ago, and they came to me themselves via DM and apologized before I’d had time and energy to put it all together and bookmark them. Case closed. I have this in my DMs, but unless they do it again I won’t be sharing their username. Fair’s fair.

If not, what would be sufficient?

Again. This is… just not really a question of interpretation. Ever seen Gone Girl? It’s like blood splatter. It’s either all over the floor or it isn’t, and while you can try to scrub it off, unless you know which chemicals to use, it’ll still stick and light up like a bitch. And there’s gotta be a goddamn lot of it for me to make a public bookmark.

Also, please, just a suggestion: consider having more people look over those works and give their unbiased opinion on whether they think it's AI or not? I believe that you don't mean harm, but basing serious accusations on one single person's analysis is not the way to go, I think.

See above, please. In some cases though, one educated person’s analysis is actually enough.

I really hope you'll answer this and not just delete my question, and I hope we can have a productive conversation about this because really I think these talks need to happen if we want to keep fandom safe for everyone. Thank you.

Thank you too, Anon, for this ask. I hope you’ve now had your questions answered and I wish you a good Sunday. 

For whoever sees this: I said I have an opinion ask in my inbox that doesn’t require me to reply. That’s cool, takes no time, into the queue. If there’s something you wanna say about this topic, you’re welcome to. Just mark it as opinion at the beginning of the ask please and I’ll post (no usernames please, general details ok).

How I wish you were right that the second AI-users need to start jumping through actual hoops they’ll give up. But guess what I’ve been observing these past few days? 

For one reason or another, some users I’ve contacted but not yet bookmarked for strong indicators of AI involvement have… seemingly mass-hidden some of their works into private ’rewrite’ collections. Not just one, but several. Is that an indication of guilt? No, not necessarily. People are allowed to edit their own works, I know authors who do it all the time. But it will be interesting to compare versions, if these works ever appear again, to see if the edited parts happen to correspond to the same things that rang my bells. We’ll see. I can’t bring myself to care too much here either. The works are hidden, so no one can stumble upon them and be misled. 

I’m sorry if this is disappointing, but I won’t be making any ’LOOK FOR THESE SIGNS TO SPOT AI’ posts because:

1: I’m not out to start some kind of mass movement. Fandom doesn’t need that. My bookmarks are at the end of the day for myself and my friends, and for cheaters to know that we see them. People are free to label me a quack, and not once have I stated anywhere that you must believe me. If someone doesn’t agree with me, let them not agree. Let them keep reading. That’s their choice.

2: Unfortunately I’ve seen too many ’this is ai’ comments (from guests and registered users, sure, some can be bots, but still) on works that are so obviously human to me ! simply because the works were extremely well-written or included the long dash. No. I’d like to say ’Don’t be an idiot, use common sense’. But common sense seems to be lacking these days, so I won’t be, even accidentally, putting any kind of encouraging indicator lists into the hands of people who consistently show they can’t be trusted with them. Is that infantilizing? Maybe. But I’d rather infantilize than say ’this is a common sign of AI’ and then see people start blindly attacking others because they lack the ability to put signs into context and weigh indicators against one another. 

This doesn’t mean though that things like the stuff I included in the past analysis won’t be mentioned here anymore. Not at all. You use the words ’educating the community’. I am no teacher, but I do have some things that I think readers could learn from. I recently contacted someone asking if I could use their work as a case study, and they’ve agreed. So I will in the future be making an ’educational post’ about what AI may do to elevate a story and show what this looks like (this has already happened, I won’t be generating anything myself). I hope you will find this informative and interesting too in due time.

Someone dropped me a DM informing me that the user in question used to go by another username. Not another account, just a different username. This username led further back in time off AO3 and revealed more (very not GPT-style writing), which changes the quantity of the comparison data. And when that happens, I need to adjust what I’ve got. And during that adjustment period, I remove the bookmark and keep the details away, because if someone were to demand an analysis right this second, it wouldn’t be complete and as you perhaps saw from a previous case, it takes a lot of time to compile everything into a package that someone who’s not GPT-acquinted can easily follow in order to make their own decision.

Side-note: I don’t consider myself or my partner or my boss or anyone an all-knowing AI God. Obviously. And as I’ve said before, I don’t care if anyone believes me. If people choose ignorance is bliss, let them. If they think I’m a hag who’s pulling shit out her ass, let them. It doesn’t hurt my ’credibility’ to be wrong when at the end of the day the only people whose opinion I care about are the ones closest to me, who know me. That also means I have no trouble apologizing if evidence suggests that’s what I should do. It would be pretty hypocritical of me to refuse to apologize when the very thing I’m asking for is a) disclosure and 2) apology if they have bullshitted the readers right in the face in the comments for example. 

I also received this just a few hours ago: 

For anyone reading: remember we do not shit on spelling errors. These are often a sign that someone has bothered to learn more than one language, which is more than can be said for half the country. And yes DM’er, I do know. I’m well aware that sometime in the days following the making of that bookmark, the author stopped doing something they used to do a lot. Is that an admission of guilt? No, people are allowed to experiment and change how they write. Of course they are. 

But we’re talking about a prevalent syntactic construction that’s been present in every single work since they began ’mass-posting’, and when people begin to change their styles, it rarely happens overnight. They’ve been publishing a perfectly-written story per day. Could they have had backlog? Sure, especially for the event I’ve wiped over. But the change is visually overnight, since I’ve got the befores and afters saved and they’re dated, and it correlates with my timing. And because I’ve got it all, it’s easy to compare post-changes, so this one’s not really at the top of my list right now. So whether I crack down on this one tomorrow or a month from now doesn’t change what’s already been uploaded and subsequently downloaded.

I’ve promised people to throw more details onto AO3 and I also need to tweak my bio (because boy do I sound pissed off) and that takes priority after asks, so I’m just not gonna stress about this one. Perhaps the author even has a change of heart in the meantime and just owns up to it. That’d really be the best outcome. 

Thanks for your nice message, Anon. A good afternoon to everyone.

Obviously I can’t make a comment about the fic, but I’ll say that if it is one of these cases then I’m sorry you’ve had the (exceedingly more common) misfortune of running into one. I was scrolling one of my favorite tags earlier on this account and found a couple of great stories to leave kudos and positive comments on. And then two cases I had to note for myself because too many bells went off like it was Christmas morning, except the gifts really sucked. You’re not alone in it. 

The reason I’ve had to wipe out part of your ask is that you’ve uncovered a clever way for AI users to cover their tracks. There are ways to check, but it’s even more time-consuming so I’m just not gonna hand it out on a silver platter on this blog of all places. But I do wanna address the rest of your ask publicly. 

What would make a fanfiction veteran take that route? I’ve had some time to wonder about this. The first thing we have to consider is that even if they are someone who’s been around for a long time and had their fics scraped, they might not care. I personally can’t imagine not caring. I can’t. But that’s me. People are different.

What else could it be? 

AI both is and isn’t a new thing. When new stuff happens, especially stuff that comes to affect society on the whole, people wanna see what all the fuss is about. Curiosity is natural. So they might wanna test it, see what it can do etc etc. Then, unfortunately, they might think hey. Nice. Because in reality it doesn’t really occur to people that what a machine gives them, it pretty much gives everyone else too. My partner and I have many times tried to force Chatty to break one of its signature codes, we gave it explicit orders, not up for interpretation. It managed for like one long paragraph out of three and by the second it had already slipped it in there. It apologized. Then it did it again. And again.

I think if someone’s a veteran, hitting a wall at some point is almost to be expected. I know it’s happened not just to me. Maybe they wanted to see if Chatty could help them creatively. That might sound like a binary opposition, AI and creativity, but it ties into the curiosity aspect. Maybe this machine can help me, I’m stuck and I need a sparring buddy. So they start, and they stay. Possibly. 

In a previous post I mentioned (without naming the AO3) a user who’s been writing forever and somewhat recently started using GPT as their beta reader. I never commented myself but someone else did, telling them it was kinda obvious, and the author replied that they were struggling to find a beta they clicked with so until then they’re using GPT. They put this into the author’s note of the first chapter. I take no issue with that whatsoever. I take my hat off. At least someone’s got a backbone in 2025. 

This came to me just now while typing out this reply. I don’t know if this has ever happened to you or if you write, but there are readers, and then there are readers. And as much as I love my readers and love getting "please update soon" comments, if the comments you start receiving are essentially just "update please", there comes a point when it stops being encouraging and becomes stressful. At least for me. Writing well is time-consuming. Maybe they feel pressured to update regularly? Maybe using AI keeps them from burning out? Or maybe they love being in fandom but can’t really bring themselves to invest as much energy and time anymore, but they don’t wanna stop, so they use AI to get some of their time back. I don’t know. These are possibilities that logically make sense to me, but emotionally, I just can’t excuse its undisclosed use for reasons I’ve mentioned many times over by now. If they use it for a specific reason, just say so.

Regarding your question whether I would burn this kind of fic? That would depend on what they said to me, if they said anything. If they owned up to it and just put in a note/tagged it, then no I wouldn’t. If they inform readers that they’re using AI in some capacity and readers choose to continue reading and leaving love etc etc, let them. We have to accept that not everyone cares about the effects of AI, no matter how messed up that sounds. When it’s what suits me personally VERSUS why AI sucks, what suits me may very well win.

I don’t think you sound like you’re ranting at all. And even if you did you are more than welcome to. This is a very ’rantable’ subject. ’Ventable’ too. Asks are open for that too. 

Be mindful out there, Anon. It’s brave of you to have spoken to this user yourself and addressed the matter. But remember what I said: some authors have huge followings and for those followings the author can do no wrong. Even when it’s plain as day. If someone’s been around for 10 years, that’s a can of worms. And I’ve seen some cans all right.