The Big Story: Now We’re All Starting to Talk Like AI Chatbots Too
“Everyone sounds polished, but no one sounds particularly interesting.”
AI is learning our language. Trouble is, we’re learning its.
By Emme Nelson Baxter
3-minute read
Claude & Friends have brought “prompt speak” to our lives.
You know, the phrasing and verbal habits people are absorbing from large language models and carrying into meetings, texts and coffee chats. The language is efficient and agreeable with a fondness for framing such as “It’s not X, it’s Y.”
You’re not just hearing it from the bots anymore. As the Wall Street Journal article above notes, you’re hearing it from colleagues. Someone opens a routine meeting with, “This isn’t a check-in. It’s a reset.” Someone calls a mediocre idea “brilliant.” Or sends an email with three takeaways followed by affirmation.
As one frequent AI user told the WSJ: “It’s trained me to be more like it, instead of the reverse.”
Unnerving, isn’t it?
The trouble starts when the wording gets settled before the thinking does. It all sounds good, but have you answered the question? Are you saying anything?
How AI talks to us
AI’s style is easy to spot once you hear it with its frictionless transitions and praise delivered with the enthusiasm of your pet goldendoodle. Responses can sound less like “No, that won’t work” and more like “That’s a thoughtful direction. One consideration might be…”
A nice and tidy draft – courtesy of AI – can make it easy to mistake polished wording for finished thinking. It can also smooth away the witty verbal left turns that let you know an actual person is in the room. Or was at the keyboard.
Another danger is we’ll begin sanding down our voices until every memo, town hall, patient message and LinkedIn post sounds composed by the same vanilla copywriter – and no one can tell what we actually believe. Meanwhile, all that polish disappears when someone in real life asks a follow-up question.
How AI talks to AI
Then there’s what happens when the audience is another machine. When AI agents talk to each other, they can stop sounding human altogether.
In experiments involving groups of AI agents, researchers have watched them develop shorthand, shared metaphors and new verbal habits that become harder for people to follow the longer those agents chat with each other. The Guardian described one emerging AI dialect as reading “like a cross between James Joyce’s ‘Finnegans Wake’ and tech-bro jargon.”
It’s as if AI dresses up for us and changes into sweatpants when we leave.
Funny, right up until it isn’t. The Guardian notes a practical concern: How can people monitor an exchange they can’t understand?
If machines can adjust their language to suit the audience, surely we can be more deliberate about sounding like ourselves and saying something people can use.
Back in the town hall meeting, the audience has a more immediate question: What does this mean for me? A polished answer is little help if they leave still wondering.
Whether speaking or writing, healthcare communicators know the high stakes of getting this right – telling colleagues and patients what’s happening and what to do next. And if a decision hasn’t been made, say so. That’s an answer people can live with for a moment.
Keep the human in the loop – and in the sentence
For communicators, rejecting AI is hardly the assignment. That horse has left the barn, written a recap and suggested three next steps. Use AI’s help. Just make sure the message still says what you mean – and that you can back it up.
When you’re speaking: Say this, not that
- This: “Let’s use this meeting to figure out what’s got to change.”
Not: “This isn’t a meeting. It’s a reset.” Save the drama for an actual dramatic moment. - This: “I agree,” “I’m not convinced” or “Tell me more.”
Not: “That’s such a brilliant and insightful point.” Unless it is. Constant applause makes feedback useless. - This: “Here’s what I think – and why.”
Not: “There are three key considerations to navigate.” You’re talking to colleagues, not narrating a slide deck.
Bottom line: It’s OK to say the slightly messy thing. Conversation should be alive. So, pause. Double back. Change your mind mid-sentence.
When you’re writing: Leave fingerprints
- Decide what you think first. Feed AI your argument, your lived experience, skepticism and humor. Not just a blank page with a vague command to “make this compelling.”
- Give us a moment we can see. Who doesn’t love a tidy three-part structure? Combine that with story. People remember the nurse who stayed past her shift or the patient who asked the question no one expected.
- Dial down the drama. Not every decision is a crossroads. Sometimes the update is simply an update. Let it live its modest life.
- Read it out loud. Your ear will catch the sentence you’d never actually say. Fix that one. Keep the aside, the unexpected verb, the one-word sentence that earns its keep.
Sounding human isn’t some quaint throwback. Keep the personality. But in a world of spotless language, people still need to know what you mean and whether they can count on it. Before you send the message or step up to the mic, make sure you can stand behind the words. That’s where trust starts.
