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

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

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

>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.

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.

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

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.

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.

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.

Did you mean "use emojis"?

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

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

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.

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:

    Anthropic Claude:

    Grok:

    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.

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.

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