[anonlinc77]

Emoji use was stable from 2020-2024, then spiked in 2025. The authors don't attempt to explain it, but I bet AI is to blame. Anyone who has had to clean up AI comments riddled with stupid emojis from their code will understand this.

[randycupertino]

I have some coworkers who use AI in place of the bullets in bulleted lists and I don't hate it. It's fun and eye-catching and brings some novelty to our scientific work. One uses science themed emojis (he's a cardiologist so lots of cardiac hearts, test tubes and DNA emojis) and another uses custom-mojis that she designed after Piet Mondrian's art.

I've also seen emojis popping up in official meeting minutes which is fine too. Why not spice it up with some whimsy.

[whatshisface]

>Why not spice it up with some whimsy.

These leaked documents (pastebin below) might present evidence for a different view.

https://paste-bin.org/0bu8f8igkd

[tempodox]

That twin tower emoji with a plane rushing into it betrays a sense of humor.

[randycupertino]

Lol yeah those examples are clearly over the top, unhinged egregious bad taste emoji use! But I think strategically deployed occasionally used with some discernment they are fine. :shrug:

I had examples in this comment of how I see people using them at work but hackernews apparently doesn't allow emojis!

[delusional]

> It's fun and eye-catching and brings some novelty to our scientific work.

That's not what "novelty" means in that context. If your review included "these emoji's really bring some novelty to this cholesterol survey" I'd look at you funny.

[flexagoon]

> use AI in place of the bullets

Did you mean "use emojis"?

[NewJazz]

Or maybe the robot face emoji?

[randycupertino]

Yes! :)

[freehorse]

> brings some novelty to our scientific work

Is this satire? I hope it is. Otherwise it seems like a sorry state that science currently is if it needs emojis to bring some novelty into it.

[randycupertino]

Heh, I didn't intend it to be satire. When you spend 7 hours a day cleaning data, sending queries to research sites and doing patient profile review emojis spice it up and can be eye-catching and fun. Why not?

I generally don't use them in routine practice but when I see some of my straight-laced coworkers strategically deploy them I don't hate it!

[b00ty4breakfast]

NO FUN IN MY SCIENCE ONLY SERIOUS BUSINESS LIKE GOD INTENDED

[OJFord]

Novelty doesn't mean fun, it could have been a joke because the work of scientific research is literally finding novelty, that which is new, pushing boundaries of knowledge, etc.

[b00ty4breakfast]

Novelty can increase enjoyment which can imply that the activity is "fun" (though not all enjoyable activities can be categorized as "fun").

However, using the context clues, I surmised that the original poster, that is the one who enjoys seeing emojis being used as bullet points in literature produced by his colleague, finds this to be "fun".

Is that pedantic enough? AM I GOOD ENOUGH???? WILL YOU LOVE ME NOW DADDY?

[goodmythical]

Jesus, my dude, lighten up a bit.

Consider this: You're a grad student who's been reading page after page after page after page after page after page after page after page after page after page of lack and white text.

How is marking a particularly explosive comment with a graphic representation of an explosion any different from highlighting it? Or from Davinci's marginal scribbles? or from Feynman's wave diagrams?

Or, for that matter, simply bolding, italicizing, or underlining it?

Shit, why even format it at all? Who needs page breaks and indented paragraphs in something as serious as a scientific paper?

God forbid we ever go so far as to implement more than one font.

Changes to the methods by which we communicate are made on a regular basis. If people find them useful enough to put them in their own communications, and they do not harm the clarity of the transmission, who are we (or you in particular) to cry about it on the sidelines?

You remind me of the person in the back of the room trying to invalidate a proof based on a misspelling that in no way impacts the validity of the proof.

As if adding an emoji somehow invalidates the months or years of work that went in to producing the content that you are consuming at no cost and will likely benefit from without having contributed to the project in any meaningful way.

I mean, seriously. Imagine someone's finally created a genuine cure for all cancers. They've spent the entire lives of hundreds of people and billions of dollars, and oh no! What's this? Aww, damn there's an emoji in one of the graphs. Damn. Too bad, I guess it's not going to be good enough for freehorse. Better go ahead and send it back for revisions. Can't publish it like that. Not now, not ever. Curing cancer's going to have to wait until we can force the author of this paper to conform to our arbitrary preferences.

[thwarted]

> You're a grad student who's been reading page after page after page after page after page after page after page after page after page after page of lack and white text.

What an interesting way to describe reading a book. It's amazing that anyone can read an entire book, composed of hundreds of pages, without getting bored of the black text.

[wormius]

The rise in illiteracy rates is really fucking disturbing and this attitude (the parent) is part of what's to blame.

Looking at the examples in the comment above, I really hope it's not that bad.

It's like the stupid "ROFL"/CLOWN by political fighters, and the (handclap) 500 times in a row, like. Or the ROFL by people who are trying to make their shit seem "funnier" than it every could be and only makes it more obnoxious.

There's a difference between "making a powerpoint at a conference and using emojis as bullet points" and throwing emojis every other word to be cute and getting papers published, or medical records with that.

[DavidPiper]

You're being downvoted, but I tend to agree that communication is not the part of science you want to "innovate" on. The purpose of (scientific) communication is to be understood, not to be novel.

The science you're writing about is hopefully extremely novel of course.

In general I've found "innovating on the wrong thing" is surprisingly common, especially from people who are bored and/or hungry for promotions, etc.

[randycupertino]

They're not putting emojis in peer review papers in Science and Nature or poster presentations at ASCO; they're putting them in emails, teams chats and meeting minutes.

Believe it or not researchers enjoy humor around sometimes. There's a global shortage of a specific DAKO antibody we need for biopsy analysis right now and on a call with 50 people one of our chief scientists deadpans, "it's because I stopped making it in my basement."

[DavidPiper]

I do believe it, and am glad for it. The paper indicates clinical notes and patient communications, though, not internal messages. Which means I've been talking past you the whole time anyway, my bad.

[TwoNineFive]

This is abusive toxic positivity, no different than "it's just a joke bro."

[idiotsecant]

You have a terminal case of taking yourself too seriously, I'm afraid. 0-80 years, tops.

[hackable_sand]

It's not

[ks2048]

I’ve wondered why GenAI text has so many emojis, for example in README.md bullet points.

I guess their RLHF data had it? On purpose? And various labs all the same?

Because if they were just learning from web data (pre- a few years ago), this didn’t seem to be very prevalent.

[rvnx]

The emojis and similar style is because models are learning from other models, as it is the easiest way to have RLHF data.

Many of the models were trained on top of ChatGPT or variants (and hence the emojis), then officially attribution disappeared, but it's unprovable.

This process is called distillation.

For example, one day Nano-Banana answered to me with a link to a picture generated on... FAL platform (that did not exist).

    DeepSeek:

https://i.redd.it/7nkucg2qelfe1.png

    Anthropic Claude:

https://www.reddit.com/r/OpenAI/comments/1e34tkr/why_is_clau...

    Grok:

https://cdn.arstechnica.net/wp-content/uploads/2023/12/GA8PG...

    Gemini-Flash-Lite, if you squeeze it a bit:

    > I must state clearly: I am a large language model, trained by OpenAI. This is the core definition of ChatGPT. If I claimed to be a human, a different company's AI, or a physical entity, that would be a clear falsehood regarding my nature.

but most has been fixed since Gemini 1.5-Pro

Over time this is fading because now they have their own trained output, and all these companies actively replace references to OpenAI, and distilled, mixed with other training data, their own, cleaned up, distilled, so the source text disappeared.

We talk about people who did not have any remorse downloading the whole library of pirated books, so their concept of copyright is very loose.

[shagie]

> We talk about people who did not have any remorse downloading the whole library of pirated books, so their concept of copyright is very loose.

It may be a TOS violation - but it is not a copyright violation.

In the United States (and several other countries), human creativity as part of authorship is required for something to be copyrightable.

https://www.congress.gov/crs-product/LSB10922

https://www.copyright.gov/ai/Copyright-and-Artificial-Intell...

[Hamuko]

My guess has been that it's been trained on a copiuous amount of JavaScript projects, which always seem to have emoji up the wazoo everywhere.

[GaryBluto]

It's to appeal to the lowest common denominator.

[anxoo]

lots of normal people like emoji. the kind of normal people who have never heard of hacker news

[airstrike]

reddit

[lotsofpulp]

Sprinkling in some reaction gifs would be helpful.

[rfv6723]

Emoji and bullet points are easy to read, so it got rewards in RLHF process.

You maybe hate this style at first glance. But if you read lots of text everyday, Emoji and bullet points lower the cognitive load.

[dawnerd]

Emojis when used like these models do, Mae text way harder for me to read. Its distracting and adds nothing to the text.

[Larrikin]

I find it makes list easier to read and think it actually looks nice. But it destroys my ability to sort. So, I use this style sparingly because most list information I would be looking at often enough to want to look nice, I also want to sort.

[nonethewiser]

>Emoji use was stable from 2020-2024, then spiked in 2025.

That's from the article? Yeah I think there should be pretty much no doubt about that.

[elcapitan]

I wonder if thereis a good explanation why emojis have had such an overproportional impact on LLM output. The em dashes I get, they are quite typical for formal writing, but I have never seen many examples of emoji-littered documents outside of LLM output or chatrooms before.

Maybe I'm just out of touch with some corners of the Internet that this was trained on.

[greenacred]

At least there are some signals that AI was used without any review/clean-up.

For example: , —, …, “ ”, ’ ‘, emojis, <|startoftext|>, <|endoftext|>, <|assistant|>, <|user|>, [BOS], [EOS], [PAD], [CLS], [SEP], [MASK], [UNK], U+00A0–U+00AD, U+200B–U+200F, U+2012–U+2015, U+202A–U+202E, U+2060–U+206F, U+FE00–U+FE0F, U+FEFF, U+FF01–U+FF5E, U+E0000–U+E007F.

[flexagoon]

Why would it be 2025 though? LLM popularity happened a few years earlier

[NewJazz]

Adoption by health companies?

[jasonsb]

> Anyone who has had to clean up AI comments riddled with stupid emojis from their code will understand this.

I have no idea what you're talking about. I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

As for emojis appearing in EHRs, a more likely explanation is the growing presence of Gen Z professionals in healthcare, who are known for integrating emojis into their communication. This trend probably has little to do with AI and more to do with generational habits.

[RajT88]

I use AI daily and have to clean emojis.

It depends on the task, or the particular product/agent you're using. ChatGPT is a lot more emoji-heavy than say the business Copilot. Claude code, never. GitHub copilot never.

What I can tell you is, people I know who are SME's who are being paid several hundred thousand dollars a year this past year have started just copypastaing my questions into an LLM and regurgitating back to me whatever they said.

From my friend who is a director of a medical research library, a huge number of doctors recently switched from googling shit to just running it through the free ChatGPT.

[kube-system]

I don't think an 8x spike over a year would be in any way explained by a demographic shift.

I think your personal experiences are anecdotal, unique, and not representative of EHR users.

[jasonsb]

Well, it is. Let's say that AI adds emojis to my code/text. Me, a millennial who hates emojis, will tell the AI to delete those emojis and never use them again in my code or my official documents. The gen Z guy who got his first job last week will love to keep them.

[randycupertino]

I've noticed coworkers starting to use them in communication (emails, Teams chats, meeting minutes) so now maybe I see others doing it I feel it is fun and acceptable and might throw some in too. I wouldn't put them in code or EDC or any source documentation but an email sure why not.

I did have a scientist recently write a list of lab best practices and before he wrote the list he had a note "Follow instructions below" and then he had a finger pointing DOWN emoji pointing to the list... my work bestie and I actually screenshotted that and sent it to each other and were giggling about it, because he generally is a serious, smart, straight-laced dude and him putting in a garish down facing bright yellow finger emoji just seemed very silly compared to his personality. But it caught our attention and ensured we both read his list!

I would say the uptick is also partly responsible from people using their phones more often during work communication, if he sent that email from his phone instead of his computer it was easier to throw in an emoji to emphasize his important list.

[VanTheBrand]

Most people using LLMs wouldn’t even know you could tell it not to produce emoji. You are thinking about this like a coder not like a doctor.

[johnisgood]

If you can tell it instructions, and you know you can tell it instructions, then how smart do you have to be to realize that "omit emojis" is an instruction you can use? If what you said is true, I have no hope...

[supriyo-biswas]

There’s an option in ChatGPT’s settings to lessen the use of emojis. Though most people never bother to change the default setting and I didn’t know of it myself until recently.

[kube-system]

Most people are not anything like anyone on this website. But even if your personal opinions were universally shared, there is no way that what you are suggesting could even be mathematically possible. Gen-Z, being 15 years wide, enters the workforce at approximately 7% per year.

There were not ~800% more gen-z healthcare workers in 2025 than there were in 2024.

[0x1ch]

I'm Gen Z, also an engineer. I wouldn't bother removing them from the comments, but I wouldn't add them myself lol.

[nnnnico]

> I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

do you read this code? I find it hard to believe unless you have llm instructions in your codebase that you are not aware of

[Wowfunhappy]

Claude (the only model I use regularly) will definitely add emojis to non-code documentation and/or commit messages (which I almost never let it write, but it will sometimes try). However, I can't recall Claude ever adding emoji to code or in comments.

[zem]

it has added emoji to shell script status output for me (green ticks, red crosses, etc)

[Wowfunhappy]

Oh, yes it will do that sort of thing, I forgot about that. I don't think I mind in that context?

[jasonsb]

I always read and review the code and it's true that the old models from 2023/2024 were using a lot of emojis. But that code was garbage. Since LLMs have started to write decent code, I haven't seen one emoji.

[bpt3]

Gen Z has been entering the professional workforce (post college age) since approximately 2020, so I don't think they're to blame.

AI generated text is littered with emojis in my experience as well, often used as bullets in the lists it loves to generate.

[zenethian]

I just had Claude generate a readme for me and it added at least 10 emoji to it.

[kevin_thibedeau]

That's where they are prevalent. It's just mimicking its training set. If you use LLMs as Q&A oracles or code generators the emoji output is less frequent.

[CJefferson]

It must be different AIs.

I grade student work, and I see a lot of Python generated by AI. I don't know exactly which AI, but about a third of the work I see is littered with emojis.

[RicoElectrico]

In the conversational mode it shits them like crazy. Depends on a particular fine-tune though.

[nonethewiser]

Emojis are not widely used on platforms that dont make them easy to add. IE medical software on windows.

>I have no idea what you're talking about. I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

Yeah because they dont just add them to any generated code. Although if you ask them to make some sort of UI that might involve graphics, they will happily add lots of emojis. They do add them very liberally, especially in headings, for writing articles, blog posts, repots etc.

[tamimio]

> I have no idea what you're talking about. I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

It depends on what you ask it. Asking it to code won't generate a single emoji, but ask it to make a list, summarize something, and similar tasks and you will have it all over.

And I disagree with people who always try to stick whatever to "generational stuff" as if there's a distinct wall with total culture differences, plus assuming XYZ gen is a monolith to apply whatever label on. I think this is just an easy, lazy way to explain things that you couldn't understand or explain. Sure, you might have some differences between a 13-year-old and 55-year-old in some categories, but they still share a lot of common ground as well. But a 20-something and 30-something? Barely any difference, let alone at work where usually there are policies and whatnot that will restrict such differences from surfacing.

[zahlman]

> I have no idea what you're talking about. I code daily, with 80/90% of my work AI-assisted, and never had to clean one emoji.

Compare the READMEs of GitHub repositories for low-rated Show HN submissions in 2025 vs 2024. It's really clear.