AI Fiction Is Easy to Detect Because It's Stupid and Bad, Research Finds

2026-07-11 02:23 UTC

Fiction written by artificial intelligence is easy to detect because it struggles with complex story structure and tends to moralize in clunky ways, according to a preprint study from researchers at University of Maryland, College Park and Google DeepMind. They found that AI fiction has tells that go beyond stereotypical overuse of em-dashes and other obvious AI tropes and have more to do with the formulaic nature of the text itself. “AI stories over-explain themes and favor tidy, single-track plots while human stories frame protagonists’ choices as more morally ambiguous and have increased temporal complexity,” the study, which looked at more than 50,000 AI-generated short stories, found. “Claude produces notably flat event escalation, GPT over-indexes on dream sequences, and Gemini defaults to external character description. We find that AI-generated stories cluster in a shared region of narrative space, while human-authored stories exhibit greater diversity. More broadly, these results suggest that differences in underlying narrative construction, not just writing style, can be used to separate human-written original works from AI-generated fiction.” Basically, AI-generated fiction sucks and at the moment is easy to detect. The typical method of detection involves looking for stylistic markers such as an abundance of em-dashes, the overuse of the word “delve,” or an obsession with goblins, but this project tried something different. “The idea for this project came because we are hoping to eventually move past plain text detection, into some sort of space where we can separate human ideas from AI-generated ideas,” Jenna Russell, a University of Maryland researcher and one of the study’s authors, told 404 Media. Russell is also an intern at the AI-detection company Pangram.

Replies (9)

@fubarx@lemmy.world 2026-07-11 03:00

🤩That’s not true! 🚀

Open ##3901191
@CubitOom@infosec.pub 2026-07-11 03:05

And yet 50 shades of grey came out before llm general availability

Open ##3901465
@queermunist@lemmy.ml 2026-07-11 04:19

Other tells I’ve noticed: lots of purple prose, sentence fragments, and the good ol “not X—Y” framing.

Open ##3907451
@the_riviera_kid@lemmy.world 2026-07-11 04:22

50 shades of grey was also stupid and bad and it very much was not written by AI.

Open ##3907672
@solrize@lemmy.ml 2026-07-11 04:57

“Research” must have never met my friend Sturgeon. (Of Sturgeon’s Law. No I didn’t actually know him).

Open ##3910237
@trem@lemmy.blahaj.zone 2026-07-11 05:42

Claude produces notably flat event escalation, GPT over-indexes on dream sequences, and Gemini defaults to external character description. Which are all ways of creating more text without saying anything. Repetition is another classic. I feel like that’s a problem with all low-effort, AI-generated texts (which is the vast majority of them): You give it a prompt with maybe ten pieces of information and expect it to generate a text that’s a hundred times as long, while also not straying too far from the prompt. Of course, it’s going to take every opportunity to not say anything that would advance the plot. Because advancing the plot means either using up the little input you gave it, or to invent new information which might contradict the prompt…

Open ##3915156
@Jordan117@lemmy.world 2026-07-11 07:02

I suspect most of this is due to imperfect RLHF that focuses on brand-safety and rhetorical tricks that get annoyingly noticeable when they’re repeated at an industrial scale. The models are approaching superhuman ability at coding and math, but primarily because they’re more easily verifiable (plus it’s where the money is). But if there were a more rigorous way to quantify what makes good fiction work, there’s no reason in principle why a strong model wouldn’t be able to “optimize” it the same way it can images or video. I’ve fine-tuned a local model on my own writing before and it’s spooky how well it can capture a stylistic voice.

Open ##3922051
@muusemuuse@sh.itjust.works 2026-07-12 05:17

Well yea, they are basically pocketFM ads on YouTube. Just awful.

Open ##3935777
@TankovayaDiviziya@lemmy.world 2026-07-11 16:10

The more we criticise AI, the more it will take notes and improve. And before we know it, we won’t be able to tell if a fictional story is man-made or AI-made. Edit: I hadn’t realise that this is FuckAI initially. I don’t love AI, I hate what it would do to arts and even the concept of jobs and livelihood, but as much as those who hate AI either don’t realise or don’t want to admit, the potential is there and the wilful ignorance and denial is really just the unspoken feeling of uncanny valley surfacing. AI is here to stay and will improve upon, unfortunately. I just wish it had been open source instead. Also, the overwhelming majority of people already can’t tell if a music is AI or not. So AI imitating art is already underway as we speak: news.sky.com/…/a-third-of-daily-music-uploads-are…

Open ##3941333