Using AI is fine; I use it daily. Posting the raw output without reading it first is the tell.
The em dashes in every other sentence. The emoji bullets in a README. A stock-feeling header image that gestures at your topic without saying anything specific about it. That opener about today's fast-paced landscape. Wikipedia maintains a public list of these patterns ("Signs of AI writing"), and a solid chunk of my feed is speedrunning it.
Here's what those patterns tell me. You saw a generated draft and posted it. You read 'leverage synergies' and hit publish. The writing isn't yours, and it looks like you never noticed.
If that's how you treat a post, I have to assume it's how you treat a pull request.
You skipped the whole job. The model hands you a draft. Your value is what you add on top: catching the bug it was confident about, throwing out the approach that looks clean but is quietly wrong.
Relay the output as is, and you've made yourself a hackathon GPT wrapper.
When Claude hands me a function, I rename its variables and hunt the edge case it missed. Then I cut half its comments.
You're the editor between the draft and the publish button.
Put the effort back in.
Anyway, enjoy the sunrise I did in Microsoft paint (no AI).
Drop the worst AI tell you've seen in comments; I'm collecting them for a Claude skill.
Top comments (38)
Everything sounds great, but I'm still curious about one thing. Why are people so fixated on em dashes these days? If you open a novel and see one, do you immediately conclude that Hemingway was secretly using ChatGPT? 😂
It's pretty much repetition to the point that people get used to seeing it. This leads to people concluding that the post is AI-Generated. It gets a bad taste due to this since the more you see it, the stronger the correlation.
Obviously, you can put dashes in your article and nothing wrong with that on paper. It's just how people are experiencing in the age of AI and it's easier to identify without going to GPTzero. Of course, that's a dangerous assumption to make if you assume it by itself, but it's something people have notice now and building up towards. Hope that makes sense Sylwia!
I completely understand that argument, and I agree that people have started associating em dashes with AI.
But here's the thing: em dashes are perfectly valid punctuation. Are we really at the point where writers should intentionally use incorrect punctuation just to avoid being accused of using AI? 😄
That feels a bit absurd to me. Otherwise we'd end up rewarding bad writing habits simply because AI happened to adopt some perfectly correct ones.
I see the concern you are stating.
Yes, em dashes are valid punctuation and it is used in a variety of articles. Here is the counter point that I learn overtime: There are many ways to write that gives the same meaning.
Everyone has a different writing style, but it can be understood (with em dashes or not). Additionally, two different sentences can mean the same thing and can be understood.
I have bad writing habits (sometimes I accidentally made a sentence past tense instead of present tense, etc), but I get recognition of my work. If someone does not understand something, they could ask me and will clarify! Nothing wrong with that and I see it sometimes on DEV. We can improve our writing skills and DEV skills!
Yes, AI is fine. Using AI to correct grammar is fine. It's just how people are feeling and this post hardens the fact about how people are feeling when seeing articles. (btw, your articles are fine since we talked about this lol).
TLDR: It's not about perfect writing. It's about if someone can understand and connect with you. That's the beauty of it. Hope this makes sense! Lmk if there is anything questions!
Sylwia absolutely Em dashes were valid punctuation long before AI existed AI learned them from us The irony is we're now accusing humans of being AI for using the same punctuation we taught the AI. That's absurd AI has certain tells Yes. But they can be spotted not guessed based on a single dash.
Here's what actually matters:
If an article is helpful, if you learn something from it if it gives you new knowledge does it really matter who wrote it? A bad article is bad whether a human wrote it or AI A good article is good regardless of the source We shouldn't dismiss everything with an em dash as AI slop We should judge the content not the punctuation.
Thanks for the conversation Sylwia. 🙌
I completely understand what you're saying, and I actually agree with most of it 🙂
I think the reason this topic sometimes annoys me is that I write grammatically correct text in my native language simply because that's how I've always written. (When I write in English, I often ask GPT to help me clean up the grammar.)
Yet every second article seems to attract at least one comment claiming it was written by AI. In one case, someone explicitly said the evidence was... em dashes. There wasn't even any other substantive argument!
@harsh2644 I completely agree with that as well 🙂
If someone has an interesting conversation with an AI, develops their own ideas, and an article comes out of that process, I think that's perfectly fine.
What people usually call "AI slop" is something entirely different. A generic article like "Python loops explained" with no original insight can be spotted from a mile away, whether it contains em dashes or not 😄
Hey Sylwia again :D
I mean can you blame them? Like I mention before, repetition they see leads to correlation. It's unfortunate, but it's just how it is currently.
I remember you talking about this when you mention that you created a different tag for the post because people think your content is AI generated and you explained that you use it to correct grammer. I still respect that to this day. Though, others may not know that, which is expected.
I found em dashes slightly annoying when reading English books, or at least I think so in hindsight. Many languages never used them before generative AI. And English doesn't really need them either. Have a look at my past DEV posts if you need a proof. These days, even Google translate will insert them, if it makes sense or not.
A German article with English em dash style is 100% AI-written, and even in English, the abundance of those dashes (plus the emojis and slop tropes) have become hard to unsee as heuristic red flags of AI smell.
P.S. I will have to stop claiming that you can't type em dashed with a physical German keyboard. You can. That's AltGr + -_ only that I never did this until today.
That is interesting...
I tend to see them occasionally before Gen AI. It's quite rare yes, but not uncommon.
@ingosteinke oh, that's interesting! One of my side hobbies is writing sci-fi, and one of my short stories will be published soon (unfortunately only in Polish 🥲).
When I write fiction, I don't let LLMs change a single line. And yet, in a 42,000-character manuscript, I counted around 10 em dashes! Not counting dialogue, of course, because those are absolutely everywhere there.
I love sf, especially the hard sf. I'd love to read it :)
It'll be published in December 😀 I'll definitely write more about it when it'll become more official 😀
Fair, I mostly agree with you, especially the convo down below. I usually shrug off one/two em-dashes on their own as just good writing; eight or nine? Definite AI.
The post wasn't truly a target against dashes. It was primarily aimed at people shipping output they obviously don't read. Punctuation is just the easiest tell-tale to spot.
I've started cutting em dashes from my own writing so I don't read as a bot. Honestly, I think there's a bigger picture here: as more companies force employees to integrate AI, will every email/thread be read the same way?
Well said and almost based too. Thanks for sharing!
Nice painting as well lol. True artwork!
I'll take the "almost based", working on the way to fully based. And thanks, I have future plans for that lil duck.
Haha, I actually have my own ways of spotting AI-generated content, and they work surprisingly often:
Em dashes are usually one of the signs, but I don't put too much weight on them because I've had a bad habit of using them long before AI-generated writing became common. What matters more is where they're placed and how frequently they're used. AI often overuses em dashes because it seems to think they automatically make writing feel more human, conversational, or thoughtful.
Another giveaway is how dramatic the opening tends to be. Don't confuse this with genuinely good human writing—great writers absolutely know how to craft a strong opening. The difference is that AI often defaults to the same exaggerated hooks and grand statements. In a world flooded with AI-generated content, those patterns start to stand out.
The number of emojis used is another one for me. I feel pretty strongly about this because choosing the right emoji while writing is actually more effort than people think. If someone publishes content frequently and every post is packed with emojis, I start to get suspicious. It's not that emojis automatically mean AI-generated slop, but seeing a wall of them over and over can feel overwhelming and exhausting.
Context drifting is another common giveaway. AI often struggles to keep the central idea of a topic consistent throughout an article. It may start with one point, wander into related ideas, and gradually lose sight of what it was originally trying to explain. I've noticed this happens less with many of Claude's models, but the trade-off is that they sometimes struggle to connect sections together smoothly. The result is content that stays on topic but can feel disjointed and harder to follow from one section to the next.
I've always loved writing. I've written a crap load of answers on Quora, unknowingly answering bot-generated questions, wrote novels that never got published, and even wrote ceremonial captions for a dedicated Michael Jackson fan account on Instagram (big MJ fan here).
I'm not against people using AI because I use it too. Most of the time, I use it to fix my grammar or soften my tone. Even though English is the main language I use at work, at home, and pretty much everywhere else, I'm still pretty bad at grammar. To be fair, I'm not much better in my native language, Khmer, either lol.
I think the reason I spot AI-generated writing fairly often is because I've spent so much time writing over the years. You naturally start noticing patterns, habits, and little quirks that keep showing up. That doesn't mean I'm always right, but there are certain things that tend to stand out once you've been around writing long enough.
I've had people accuse me of using AI for all of my content before, and honestly, I don't really care. I use AI, sure, but then again mostly for grammar fixes and toning down my wording.
If you've got something to say about the actual topic, I'm happy to chat. If the entire discussion is just "this is AI," it's usually not worth my time lol.
You're right about the signs and I'd add one more: short paragraphs ending with a period on their own line. That's another tell But here's what I've learned after months of writing with AI
100% AI = 0% soul The knowledge might be there, but the human touch isn't.
100% human = hard to scale.
The sweet spot I've found 90% me, 10% AI I write the ideas the feelings the messy human parts AI helps me structure, rephrase, and research The voice stays mine.
AI slop is boring AI-assisted human-led writing? That's just good writing with better tools.
Thanks for calling this out We need more honesty about what's actually valuable. 🙌
Good call on the one-line paragraphs, I'll add that to my list. 90/10 sounds about right. The ideas have to be yours; otherwise, there's nothing holding the context together.
I think the high value lies in cutting all the extra fluff it generates. I had it generate some architecture docs and ended up cutting 1600 lines to 650.
The only other option is using: , ; as is that and or so in replacing em dashes!
You tell the ai not use em dashes—it goes back to using them, even when I have specifically advised in memory to not use em dashes ever...
Ai just loves em dashes Lol
I do make 50% of the intros to my articles human writing; the technical pieces would take weeks of writing, which is just not feasible.
Everything in life is a compromise, quality to quantity. In the past, people spent their time hunting for more content; now there is too much, but the quality is low.
Unlike the New York Times, we don't have a floor of fact-checkers and editors getting paid $100K a year to make 1 article a week.
Seems even the NY Times is now using ai:theatlantic.com/culture/2026/03/ho...
Go back to reading paper books pre-2020? My opportunity charity shop has books stacked up for a giant bag for $15, and guess what? They can't get rid of them.
The memory thing is huge. I've tried system prompts, memories, and skills, and it still manages to sneak a few dashes in. I have a suspicion that they've been fused into the weights or something.
Doing intros by hand makes sense. That's usually where your voice has to breathe. Although I won't lie and say I've never used a few LinkedIn optimizer bots on my intro paragraphs.
When you want to escape the AI slop, go back and read these magazines from the 90's on the Internet Archive. True cyber nostalgia, all made organically. So much fun seeing what they thought technology was going to be like.
I was going to create a vault series updating these articles in Mondo magazine, comparing what they got right to today, if there is interest, I might weave it into a larger technology article.
archive.org/details/Mondo.2000.Iss...
Great writing!!
Did you know big AI companies are scraping data from Google, Reddit, and X from 2022? About 60% of data on the internet is slop/generated.
Companies like Anthropic and OpenAI are investing billions to get that data, because it was human-produced.
Example:
If you prompt AI to generate an app, it generates the same dashboard, UI gradients, chunky components which is totally slop!! Unless you give AI a Design.md (which is also generated by AI 😭)
To be honest, I'm tired of this slop...!!
YES! I just attended a local DEFCON group meeting, and we talked about how AI-detection methods are going to get worse and worse as
The internet is going to context-rot itself away
The design.md thing is huge. I'm so tired of every other webpage and startup being ShadCN with the same gradients. Bring back originality!!😓
😭😭🙏
When I use AI to help me flesh out a blog post or forum post, I start with the outline / rough draft that I wrote, usually have the AI make some passes on it in the middle to help identify things that don't make sense, suggest improvements, and once it's done I do the tail end to make sure I add 'more personality'. I never just say "write me a blog post that talks about X". That's aggravating because they always read the same.
The tells for me are the "AI-isms" em-dash, 1.2.3. format, "Honestly, ..." and so on. But also a lack of personality and humanness. I always try to have one joke (usually bad) and something borderline edgy when I write so it stands out, but also reflects me as a person.
This seems to be the ideal workflow. Outline -> let it poke holes, then do the final pass to add back in the personality/spark.
I like the bad-joke+edgy ness. Great way to add the human back into the equation.
In software, we don't ship generated code without review. It's interesting that many people still publish generated writing without applying the same standard.
em dashes are a bad tell at this point - been using them since 2015. the real catch is asking someone to explain a specific claim from the piece.
the "I assume it's how you treat a pull request" line is the real thesis here. same pattern in code review — when someone ships AI output unchanged, every variable is
resultordataortemp, and there's always one edge case with an obvious null that the model assumed would never happen.the 1600 to 650 line cut on architecture docs is a useful ratio. the model front loads context you already have because it can't know what you know. half the tokens are scaffolding for its own uncertainty, not for you.
do you find the overexplaining pattern worse on recent models, or has it gotten cleaner?
Some comments may only be visible to logged-in visitors. Sign in to view all comments.