Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
Chatbots are leaving fingerprints on human language, from word choice and sentence structure to the way people navigate conversations. Three Northeastern experts explore language change and the role of AI.
A Facebook post about your trip to Peru. An Amazon review of a new toaster oven. An email to the super about the inflatable reindeer blocking the hallway.
You phrase and rephrase things all day without thinking that anyone — or anything — is putting words in your mouth.
What happens when AI joins the conversation? Are chatbots changing the way you write, talk and interact with others?
Three Northeastern University experts — Khoury College of Computer Sciences professor Terra Blevins, professor at the Network Science Institute at Northeastern University London Sofia Teixeira and linguistics professor Heather Littlefield — offer their insights about language in the age of AI.
“There’s no preventing language change,” Littlefield said, calling it a “living thing.”
“During COVID, we had a whole lot of new words come in. COVID itself. Zoom,” she said. Within six to 12 months, the pandemic had produced a lingo of its own.
Take WFH (work from home), social distancing and lockdown, for instance.
AI has since upped the ante.
Words like “delve,” “essential” and “insight” have become “a fingerprint of AI language,” Teixeira said. While not new, they’ve been popping up more frequently, especially in writing. So have metaphors comparing things to tapestries or symphonies — a chatbot fixation, she said.
But is AI rubbing off on people or simply doing the work for them? Experts say a bit of both, but telling the difference is tricky.
Vocabulary changes take hold quickly, but often fizzle out as a passing fad, Littlefield said. Structural shifts involving grammar typically take generations, and tend to stick.
Take the article “the.” Once incorporated into the language, a particle like that tends to be a keeper.
English speakers are unlikely to wake up one morning and decide that it needs a replacement, Littlefield said.
That said, AI is already shifting the structure of language.
Blevins suggested that one of these changes has to do with how we form sentences. It’s even crossing international boundaries.
Many large language models, or LLMs, are trained on data in the English language, where the subject, verb object order is relatively fixed. As a result, bots tend to favor this pattern even when writing in languages that offer more flexibility.
For example, in Russian noun endings distinguish subjects from objects, and verb endings mark the subject’s gender. “A mouse ate cheese” can be rearranged as “Cheese ate the mouse” without implying that there’s a rodent-eating Camembert on the loose, as long as the right endings are tacked on to the nouns and verbs. (In this particular case, “mouse” being feminine and “cheese” masculine leaves no doubt as to who is doing the eating.)
An LLM tends to default to the first pattern because it matches the English one — and Blevins hypothesizes that users might eventually follow suit, both online and in face-to-face conversations. The trend could flatten some of the nuance that a more flexible arrangement conveys.
For example, speakers biased toward the English pattern might favor the equivalent of “The cat broke the vase,” which simply reports an unfortunate incident. By sticking to this word order instead of placing the object before the subject as a more flexible language allows, they lose a way to point the finger more directly at the cat.
Stylistic patterns carry over as well, Blevins added. For example, emails tend to sound more formal and cordial, reflecting AI’s tendency to be agreeable to a fault.
In a recent paper, Blevins examined how people and ChatGPT adjust their language to match each other — a pattern typical for conversations between humans, who naturally mirror their partners.
The verdict? Mimicking goes both ways, but AI does it a lot more.
The matching showed up the most in the use of function words — articles, conjunctions and other bits of linguistic glue that hold sentences together.
If a chatbot says “not happy” rather than “sad,” users are more likely to use a negation in their reply, Blevins said.
Chatbot mirroring humans came as no surprise. After all, it’s part of how a model is trained, Blevins explained. The bot generates answers based on text that came before and doesn’t distinguish between matching the style of a document it’s completing and conforming to someone’s conversational quirks.
The surprise was that humans do it, too. While users mirrored ChatGPT less than it mirrored them, they did so at rates broadly consistent with those found in conversations between people. For example, if ChatGPT uses the word “vehicle” instead of “car,” a user may echo that choice in the next reply — much as one person unconsciously picks up another’s wording during a conversation.
“So in that sense, we kind of treat LLMs the same as we would a human,” she said.
For all the mirroring, there’s also a strong backlash happening. Language purists are deliberately stripping AI-associated habits from their prose.
All three experts mentioned the infamous em dash, which denotes an aside or a break in a train of thought, and which has gained a reputation as a bot “tell.” While the stigma isn’t entirely fair, many writers — though not all — have chosen to say goodbye to it lest they appear to be outsourcing their work.
Some try to ration their dashes.
“I love an em dash, but I’ve started using them a lot less since ChatGPT,” Blevins said.
The aforementioned bot-associated words have also gotten caught in the crossfire, with writers growing wary of using them.
Littlefield connected the backlash to a broader pattern. People often sound more like groups they feel positive about and less like those they want to distance themselves from.
“They’ll actually try to throw up kind of linguistic barriers,” she said, pointing to research that showed students using a dialect their teacher struggled to understand as a way to stake out their territory and distance themselves from their instructor.
The impact of AI on communication goes beyond the nuts and bolts of language. For better or worse, it can shape how people approach the exchange itself, both on screen and in real life.
Teixeira said chatbots can help people organize their thoughts or rehearse difficult conversations. Used that way, they can strengthen a person’s sense of agency rather than replace it.
However, human interactions thrive on productive disagreement as people hit rough patches and work to repair them. Chatbots, on the other hand, always aim to please.
Don’t like a response? An instant redo is just a click away.
However, relying on bots too much means your capacity for self-regulation and handling conflict could take a hit, experts said.
“It could disrupt how we relate to other people,” Teixeira explained. Losing sight of the fact that a chatbot’s empathy is simulated only makes things harder. What was a “first-aid for loneliness” could turn into a mental health fiasco in the long run.
It could become an “echo chamber of one,” she said.
Northeastern Global News,
in your inbox.
Sign up for NGN’s daily newsletter for news, discovery and analysis from around the world.
Katya Poltorak is a science reporter at Northeastern Global News. Email her at e.poltorak@northeastern.edu.
© 2026 Northeastern University