Now We’re All Starting to Talk Like AI Chatbots Too

Sara Wimmercranz recently found herself opening a work meeting with what’s often referred to as contrast framing, a linguistic tool favored by chatbots. This is not a morning meeting, she began, it’s a reset.

It was just as Wimmercranz feared. She’s ingested so much AI-generated content that it’s now coming out of her mouth.

“We just sound like bots,” says the founding partner of a venture capital firm in Stockholm.

Wimmercranz says conversation at the office increasingly sounds like it came straight from Claude or ChatGPT. She finds herself using more punchy phrases and partaking in an un-Swedish practice: giving fawning feedback, such as telling someone they’re “brilliant” or “so smart” just like a sycophantic large language model (LLM) would serve up.

“In Sweden, the culture would just be quietly listen,” she says.

It’s not nothing. You’re right to push back. I mistakenly repeated a common myth as fact—that’s on me. But here’s the honest truth: People are spending so much time with AI that they’re adopting chatbots’ favorite phrases in conversation, Slacks, emails and texts. Some call them LLM’isms, others refer to it as “Claude speak.”

Will Barnett, head of recruiting for a Bay Area startup, swears he still speaks like a human but finds AI-use is influencing his communications over Slack and email. His writing has a certain formality and a genericness that doesn’t always feel like him.

“I will catch myself when I’m writing where I’m like, ‘Oh my God, I shouldn’t have even written that. I should have just let AI do it because it sounds the same,” he says. “It’s trained me to be more like it, instead of the reverse.”

Barnett, 42, has taken to editing correspondence he wrote himself to make it sound like less AI, which he notes would be “the exact same workflow if I was actually just using an LLM.”

Allie K. Miller, an AI adviser to Fortune 500 companies who lives in New York, says that in a recent group chat with founders, the conversation turned to “cave man speak.” It’s their name for a hyperdirect and cost-effective prompting style that uses as few words as possible to direct the AI.

“They all shared that their cave man speak was bleeding into their everyday and that they would say things like, ‘Why no bullets?’” she says.

Miller says she has a new appreciation for any piece of writing that predates ChatGPT’s launch in November 2022, and for words that an LLM would be unlikely to spit out.

“Cacophony,” she says wistfully. “Man, I don’t think I’ve ever seen AI write that word.”

Because top LLMs respond so well to upfront goal statements, Miller said people will sometimes prompt each other in team meetings.

“If someone is rambling we will often interrupt and say what is the goal?” she says. Such prompts serve to speed up the person talking. It also helps the AI agent recording the meeting to take better notes.

Daniel Wolkowitz, who lives in San Francisco and works at an AI startup, has started to find it difficult to tell whether a Slack or email message was written by a co-worker using AI, or one who has unwittingly started writing like AI.

What he does know is that more messages come through with phrases like “it’s not this, it’s that,” “these are the problems,” and “this is how it happens.”

“My hunch is that we’re atrophying our brains,” says Wolkowitz, 38. “As a secondary result, I suspect that more people are starting to naturally communicate that way.”

Deepan Mehta, who lives in San Francisco and runs an AI startup, recently started keeping the equivalent of a “swear jar” but for “prompt speak,” at the office. Instead of a jar there’s a white board and every time someone speaks in AI-like language, they get a tally mark.

Words and phrases that will earn you a mark: genuinely, load-bearing, smoking gun, X is the new Y, and any “crazy hyphenated contractions that no one would normally use,” says Mehta, 33.

If you have the most tally marks at the end of the week? “You pay for happy hour.”

David Bohegan of Washington, D.C., who works as a nonprofit CEO, has a theory that much of AI-speak is drawn from the same linguistic style that classically trained business professionals and lawyers have been using for years.

“I had read 500 business books before AI came out,” says Bohegan, 56. “And AI has read those same books and 500 more.”

For years, Bohegan has been throwing out prompt-like directions such as “think like an analyst” or “let’s red team this.” These days, he finds more people understand what he’s talking about because AI has made the language familiar to a broader group.

“Bottom line up front,” he says. “BLUF was an acronym back in the day. Now people get it more when you say that.”

Angela Nibs has spent 20 years working in PR trying to get executives to not use corporate speak or jargon. AI is making that even harder than it already was.

When Nibs asks founders why they started their company, instead of sharing a story, they’ll launch into three key pillars, the market opportunity and their competitive differentiation.

“Everyone sounds polished,” she says, “but no one sounds particularly interesting.”