Just in the past 18 months, AI has been silently reshaping how we communicate our ideas, perceptions and feelings, a trend that linguists say is quietly influencing our culture in ways we don’t yet fully acknowledge nor understand.
NEW YORK—Delve. Adept. Tapestry. Realm, Meticulous and Underscore.
No, these are not some of the words featured in this year’s 100th annual Scripps National Spelling Bee competition.
They are, rather, some of the words now widely considered to be AI-speak, words that AI is making us use far more frequently in our written communications, and now also are influencing how we speak—especially in the past 18 months, ever since ChatGPT became the fastest-growing consumer app in history.
A new study by a group of researchers analyzing more than 350,000 YouTube videos and 771,000 podcast episodes—from before and after ChatGPT’s release in 2022—cites a massive surge of AI-generated words now flooding our academic papers, business reports and conversations. It’s not just those words used in podcasts and scripted videos, but people are actually starting to talk like ChatGPT.
“The patterns that are stored in AI technology seem to be transmitting back to the human mind,” says Levin Brinkmann, a lead researcher and co-author of the study by the Max Planck Institute for Human Development in Berlin. “This is creating a new feedback mechanism between AI and humans.”
Brinkmann’s associate, researcher Hiromu Yakura, said in a recent interview that “word frequency can shape our discourse or arguments about situations. That carries the possibility of changing our culture.”
Sound a little creepy? I am both terrified and extremely intrigued.
So is etymologist and linguist Adam Aleksic, author of Algospeak, How Social Media is Transforming the Future of Language—an exploration of how internet algorithms are influencing how we communicate in unprecedented ways. “You can actually see that between 2022 and 2024, there was a 1400% increase in our use of the word ‘delve,’” he says—one of those words most disproportionately used by ChatGPT. “And if you use the em-dash too much, you’re not just speaking like ChatGPT but you’re writing like ChatGPT, which really sucks because people who have been using it for years [like me] are now more likely to get accused of sounding like an algorithm.”
Tech Talk
This isn’t the first time our language has been re-routed by tech. From our first widespread use of “hello” in the 1870s to spellcheck in the 1980s, technology has shaped how we connect with each other.
According to the Oxford English Dictionary, the word “hello” didn’t surge into common use until the telephone arrived. The first telephone books recommended its use as the correct way to start a phone conversation. We now take autocorrect and spellcheck for granted, and many of us no longer think twice about letting Google finish a sentence when we’re taking fast notes.
And now, says Aleksic, young people, especially Gen Zers and Alphas, say ‘let me ask Chat’ in much the same way as we all would say “let’s Google it.” But what’s different about AI, Brinkmann and others suggest, is that many of us are starting to treat AI as a single human friend, forgetting it’s just a tool. “Our perception that ChatGPT is kind of like a human is affecting the way we communicate to it and with it, and to each other.”
Should we be concerned? Yes and no, say linguists.
“When these tools are used, not for cognition but guidance and editing, they can be very useful for helping us find new ways to express ourselves,” Aleksic says.
“I think maybe the concern about Algospeak should come in when we let AI think for us. AI is a new canvas to use, but we shouldn’t let that canvas dictate what kind of art we make.”
Cultural Backlash?
So perhaps it should come as no surprise that there has been, over the past year, a new kind of cultural backlash building up against AI-speak. Here are some of the trends reflecting ways some people are routing their language around the influence of ChatGPT and other large language models:
Use of slang abbreviations. These were starting to be popularized earlier this year to duck the use of words that could be tracked by algorithms and embedded into more common usage. Consider the rising use of ‘icl ts pmo’ in texting and social media —which stands for, “I can’t like this sh—. It pisses me off.” Aleksic says such abbreviations “are something only humans can understand, and once the machines catch up, the meme dies.”
Avoidance speech. “At this point, ChatGPT is so culturally present now that we are using avoidance speech to avoid sounding like ChatGPT or other large language models (LLMs),” says Brinkmann. “A lot of people don’t use ‘delve’ anymore, nor the em dash, just so they can’t be accused of having an algorithm speak for them.” On the other hand, adds Aleksic, “words not yet known to stick out as much are still worming their way into our vocabulary without us really noticing it or being able to identify it as yet.”
Prompt speak. According to Taylor Lorenz, author of the User Mag newsletter on Substack, people protesting AI-speak now sometimes communicate with each other as they would to ChatGPT, to prompt certain types of humorous responses. “A lot of the slop on Instagram right now is joking, playing into the idea that not all speech should be AI-generated,” she says. In Italy, younger generations have begun using AI-generated images, but not without obvious human inputs, like a shark wearing sneakers. “It’s an absurd new genre of Italian illustration and caricature that’s trending on social media, mostly to reclaim power over AI, making a statement about the use of algorithms and what’s acceptable.”
What’s Next?
Mor Naaman, a professor of information science at Cornell Tech, says it’s important to start thinking about distinguishing between a structured message and a personalized one, now seen as being the only kind of authentic and sincere communication some will accept.
“We have to be careful not to put our personhood at stake in this era of AI’s influence over how we communicate” says Naaman, who leads a research group looking at topics at the intersection of tech, media and democracy. “Instead of adopting a homogeneous vocabulary and a ‘correct’ method of communication, we must self-regulate and celebrate the imperfections that build trust and make us human.”
What words might you be avoiding because they’re so commonly used by AI? Share it here, along with your take on how AI is influencing how we talk, write and communicate with one another. As always, thanks for your readership!
You can create sops to remove ai-isms from your text. I'm still so sad about the em dash