alt Hacker NewsLLM Writing Tropes.md
Comments
If you are serious about sharing written ideas, I suggest you avoid using this type of prompts at all cost. I've worked with LLMs to write on my blog and they are pretty good at first glance [0]. But do it a few time and you'll notice that those tropes are the least of your problems. Not only all your articles will sound the same, but you'll see that same voice on other blogs, news articles, white paper, etc. It's as if they were all written by Mo Samuels. Readers are often here for the author's voice, not just the content of the text.
I often hear this here: "if you don't bother writing, why should I bother reading?" In fact, save us some time and just share the prompt.
I work on research studying LLM writing styles, so I am going to have to steal this. I've seen plenty of lists of LLM style features, but this is the first one I noticed that mentions "tapestry", which we found is GPT-4o's second-most-overused word (after "camaraderie", for some reason).[1] We used a set of grammatical features in our initial style comparisons (like present participles, which GPT-4o loved so much that they were a pretty accurate classifier on their own), but it shouldn't be too hard to pattern-match some of these other features and quantify them.
If anyone who works on LLMs is reading, a question: When we've tried base models (no instruction tuning/RLHF, just text completion), they show far fewer stylistic anomalies like this. So it's not that the training data is weird. It's something in instruction-tuning that's doing it. Do you ask the human raters to evaluate style? Is there a rubric? Why is the instruction tuning pushing such a noticeable style shift?
[1] https://www.pnas.org/doi/10.1073/pnas.2422455122, preprint at https://arxiv.org/abs/2410.16107. Working on extending this to more recent models and other grammatical features now
Some of these are neccesary parts of LLM's. They use the content they create to direct what they are going to say. This applies to patterns like "In conclusion, ..." and what the author calls "Fractal summary". Turn them off, and the general quality of the AI thought gets lower.
AI writing sucks because it doesn't have a voice. It's not trying to say anything. Human writers are interesting because they offer a unique perspective from their lived experience.
It also struggles to maintain deep coherence. This is all probably related. It might be very hard or impossible to have deep coherence without human-like goals, memory, or sense of self.
I feel like the audience of the file is more for me the reader rather than the LLM.
> Add this file to your AI assistant's system prompt or context to help it avoid common AI writing patterns.
So if I put this into my LLM's conversation it is like I am instructing it to put this into its AI assistant's system prompt, so the AI assistant's AI assistant.
The alternative is to say:
"Here is a list of common AI tropes for you to avoid"
All tropes are described for me to understand what that AIs do wrong:
> Overuse of "quietly" and similar adverbs to convey subtle importance or understated power.
But this in fact instructs the assistant to start overusing the word 'quietly' rather than stop overusing it.
This is then counteracted a bit with the 'avoid the following...' but this means the file is full of contradictions.
Instead you'd need to say:
"Don't overuse 'quietly', use ... instead"
So while this is a great idea and list, I feel the execution is muddled by the explanation of what it is. I'd separate the presentation to us the user of assistants and the intended consumer, the actual assistants.
I've had claude rewrite it and put it in this gist:
https://gist.github.com/abuisman/05c766310cae4725914cd414639...
These things are always so misguided, and this was no exception. The only way to have a piece of writing not flagged as AI is to write poorly. Ignore grammar, misspell words, etc. Don't follow basic guidance on composition. Generally write in such a way that you would merit no better than a C on a high school writing task.
I'll give some examples. Some will be from this list of "AI writing tropes" and some will be from prominent human-written (prior to 2020) sources. Guess which is which (answer at the bottom).
- "Let's explore this idea further."
- "workload creep"
- "Navigating the complex landscape of "
- "Let's delve into the details"
And I'm not going to get into how silly this is as a so-called LLM trope: "Every bullet point or list item starts with a bolded phrase or sentence." I remember reading paperbacks published before the first PC that used this style.
Fractal summaries is literally how composition is taught to students. Avoiding that style will make the writing more likely to sound less like a person wrote it.
I would suggest the author upgrade this to a modern version of Strunk & White and go on a mission to sell that. Call it Anti-Corpspeak or whatever. But don't pretend that these formulations only arrived in bulk in the last 2-3 years.
ANSWER KEY: these are all obviously prominent in text published before LLMs hit, as well as in the tropes doc. They are no more signs of LLM-generated text than is the practice of using nouns, verbs, and adjectives to convey ideas.
I tried using Gemini for some light historical research. It could not stop using tech metaphors. Lords were the CEOs of their time, pope was the most important influencer, vassal uprisings were job interviews, etc. The metaphors were almost comically useless and imprecise, and Gemini kept using them even when I explicitly asked it to not do that.
No mention of Claude/ChatGPT's favourite new word genuine and friends? They also like using real and honest when giving advice. Far as I can tell this is a new-ish change.
> Honestly? We should address X first. It's a genuine issue and we've found a real bug here.
Honorable mention: "no <thing you told me not to do>". I guess this helps reassure adherence to the prompt? I see that one all the time in vibe coded PRs.
A substantive human-written resource: https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
Wikipedia also has an exhaustive guide, though it's not fun finding tropes you use yourself (I'm very guilty of the false range "from X to Y" thing):
https://en.wikipedia.org/wiki/Wikipedia:Signs_of_AI_writing
Another one that seems impossible for LLMs to avoid: breaking article into a title and a subtitle, separated by a colon. Even if you explicitly tell it not to, it'll do it.
If this bugs you, open chatGPT personality settings, choose “efficient” base style, and turn off the enthusiasm and warmth sliders
It makes a tremendous difference. Almost everything on this list is the emotional fluff ChatGPT injects to simulate a personality.
Weirdly, LLMs seem to break with these instructions. They simply ignore them, almost as if the pretraining/RL weights are so heavy, no amount of system prompting can override it
A subtle tell for generated text is just how damn flat it is to read. Not that technical documentation require some form of grand prose, but how unspecific the text can truly get. Reading a high school persuasive essay can have more detail, and those are often just written for a grade.
I can understand someone needing help with writing but getting an agent to do the job for you feels like a personal defeat.
>Disclaimer: Creation of this file was AI-assisted. If you thought I was going to write out a .md file for AI myself you must be mad. AI for AI. Human for Human.
'you must be mad'. Aggressively hilarious. Love it!
The funniest thing about this list is that it exists at all. We now have a cottage industry of people trying to sand the fingerprints off AI-generated text instead of just... writing the thing themselves. If your idea is good enough to share, it's good enough to spend 20 minutes actually writing. If it's not, a cleaner prompt isn't going to save it.
Would be interesting to turn this into code (or an external model call) that can check any writing, so instead of just handing it to an LLM and hoping the LLM obeys, a set of checks has to pass before the LLM’s writing is even shown to a human..
Kind of like enforcing linting or pre-commit checks but for prose.
"The "It's not X -- it's Y" pattern, often with an em dash. The single most commonly identified AI writing tell. Man I f*cking hate it. AI uses this to create false profundity by framing everything as a surprising reframe. One in a piece can be effective; ten in a blog post is a genuine insult to the reader. Before LLMs, people simply did not write like this at scale."
This one hit home... the first time I ever saw Claude do it I really liked it. It's amazing how quickly it became the #1 most aggravating thing it does just through sheer overuse. And of course now it's rampant in writing everywhere.
Sometimes the elements in the `Avoid patterns like:` list are quoted text it should avoid, sometimes they're just descriptions of things to avoid. But they are "quoted" in both cases which is a bit confusing. Maybe not to an AI though.
It's unfortunate that "smart" quotes are listed. Most CMS worth their salt (and even static site generators and word processors) should be able to generate typographically appropriate and localized quotes for anything that isn't a quick comment.
This contains a lot of advice about good writing in general. Ironically I’d recommend it to humans as well as AIs.
Great list. Invented Concept Labels is the one I think I get most frustrated by. When exploring new areas, I’ll read its paragraphs of acronyms and weird words and think I just don’t know some term of art, and as soon as I ask for a definition it’s like, “I just made that up, that’s not a formal term, blah blah blah.”
As the article points out at the end, these aren't bad per se. The issue is that LLMs overuse them, and we're all getting the same(-ish) LLM. It's not so different from how people sometimes have their idiosyncratic phrasings they use all the time.
Great list. The one that grates most for me not on this list is aggressive use of first-person-plural and second-person perspective:
“We’ve all been there.”
“Your first instinct might be…”
“Now you have a…”
The article has been slashdotted so I don't know if this one is in there but:
One I've seen Gemini using a lot is the "I'll shoot straight with you" preamble (or similar phrasing), when it's about to tell me it can't answer the question.
What we really need is a browser plugin underlining these patterns, especially for comments.
I sort of think the whole middlebrow angst thread about Bourdieu going on right now applies to LLM writing
Interestingly, one of the claimed references is a broken link: https://www.theregister.com/2026/02/16/ai_writing_semantic_a.... It's actually: https://www.theregister.com/2026/02/16/semantic_ablation_ai_....
Trope -> Snowclone
This seems like it would have adversarial consequences. Wouldn't these list of tropes get longer over the years.
Reading through this i feel like i'm on substack
They need to add “comprehensive tests” for Claude.
Another trope: longer README.md's than anyone would make, or want.
Don't forget "The Ludlum Delusion"- every header in an article or readme reads like a Robert Ludlum novel title, ie "The [Noun:0.9|Adjective:0.1] [Noun]".
But please tell me "Why it matters" - LLMs seem to love adding this
Can someone explain why LLM's write like this when most humans don't?
This makes me think of the attractiveness of overly bad writing to writers, as a challenge, the most obvious example being the bulwer-lytton award, or the instinctive ignoring of instructions from fiction magazines that might say "we don't want any stories about murderous grandparents, French bashing, bestiality, bank robbers from the future, or kind-hearted Nazis - and especially do not try to be super brilliant and funny and send us your story about kind-hearted Nazi bank-robbing french-bashing grandparents that like killing people and having sexy fun times with barnyard animals! Because every original thinker like you thinks they are the first to have come up with that idea!" and then as a writer you feel challenged to do exactly what they say they don't want because what a glorious triumph if you manage to outdo everyone and get your dreck published because it's dreck that is so bad it's good!
It does not seem like there are lots of people who are perversely inclined to write a story with all these tropes and words in it, but surely there must be some, because if you make something that beats the LLM (by being creatively good) using all the crap the LLM uses, it would seem some sort of John Henry triumph (discounting the final end of John Henry of course, which is a real downer)
Its so sad that perfectly fine patterns of writing are now associated with slop. Just because corporate greed couldnt stop themselves from making a bubble they then had to shove down our throats to prevent popping
Honestly, you need a tailored one of these for each of the major LLM model/version pairs. Claude and Gemini don't exhibit all of the same tropes in the same severities as OpenAI's GPT series, and within each of those, each revision sometimes exhibits substantial variance from the stylistic propensities of its immediate predecessor.
alternatively: "a guide for humans on to how to sound like LLMs"
Can we do this for everyone?
It's a bold strategy cotton. Bold of you to say that. Wild how mundane things get call wild. Thay're making calling things wild their entire personality. In that case, by your logic, (least generous misrepresentation of your logic).
Looks like this was a Show HN that didn't get much attention:
Show HN: Tropes.fyi – Name and shame AI writing - https://news.ycombinator.com/item?id=47088813 - Feb 2026 (3 comments)
I hope ossa-ma sees this second round!
If only we could fix how it writes like garbage
One annoying trope I keep seeing in Gemini output is the punchy invented concept name in a tripartite list:
- “The Pledge”:…
- “The Turn”:…
- “The Prestige”:…
(For this particular example I used real terms from the stage magic world, at least according to Christopher Nolan’s film, as it captures the same meaningless-to-the-uninitiated quality.)
Many of these are standard fare in legal writing.
Negative parallelism is a staple of briefs. "This case is not about free speech. It is about fraud." It does real work when you're contesting the other side's framing.
Tricolons and anaphora are used as persuasion techniques for closing arguments and appellate briefs.
Short punchy fragments help in persuasive briefs where judges are skimming. "The statute is unambiguous."
As with the em dash - let's not throw the baby out with the bath water.
I find this repellent; why not, instead of trying to push unwelcome generated prose below the radar, stop trying to waste everyone's time? People don't object to these patterns because they hate lists of three; they object to them in this context because of what they signal about the content.
If using AI to write is nothing to be ashamed of, then you shouldn't feel the need to hide it. If it is something to be ashamed of, then you should stop doing it. If someone objects to you poisoning a well, the correct response is not to use a more subtle poison.