Parasocial Relationships With AI Inspire Cambridge Dictionary Word Of The Year — And Warnings From Experts

The Cambridge Dictionary 2025 Word of the Year is “parasocial,” a term long used to describe the one-sided relationship between an obsessed fan and a celebrity, but which is now used to describe one-sided relationships with generative artificial intelligence (AI) tools like chatbots.

On June 30, 2025, searches for the word “parasocial” spiked when YouTuber IShowSpeed blocked a fan who described themselves as his “number 1 parasocial,” but Cambridge Dictionary said searches for the term have been steadily increasing, driven in large part by “media coverage about Meta and OpenAI and the potential effect of their chatbots on children and mental health.”

In September, Cambridge officially updated the definition of parasocial to include one-sided relationships with an AI.

Pediatric Psychologist Dr. Erin Newins, host of the PBS sitcom “How Are We Today?” says there can be some benefits to parasocial relationships with celebrities or influencers, such as exposure to a community of like-minded fans, which can help “reduce isolation.” She warns, however, that caution is warranted when it comes to using chatbots.

“Weve been having this conversation with kids for years, navigating the difference between real-life friendships and online friendships. This is just another facet of that: Does relating to or communicating with online personalities meet the need for social interaction and connection? Id likely say no because this is missing the reciprocity, but at the end of the day, Id rather someone have at least SOME social interaction, an online internet connection, or involvement in a group is better than none,” she says.

“Overall, I think having a connection to social communities or fan bases and having favorite celebrities or influences can be positive if it includes an effective understanding of boundaries and isnt a replacement for meaningful reciprocal connection,” she says. “In general, I would be cautious about putting any energy into a connection with any bot.”

Dr. Toby Watson, a licensed clinical psychologist and Director of Associated Psychological Health Services, an outpatient practice in Wisconsin where he has treated children, adults, and families for more than 20 years, says generative AI bots can be beneficial “when used properly,” but much like psychotherapy, emotional dependency can develop, “especially when it is immediately available 24/7.”

“AI bots can take hours and hours with a user to practice social skills, communication skills, and role play till the cows come home. Taken together, and in the case of GROK ‘Therapist’ AI bot, you literally have an expert therapist-like companion who is available 24/7, who never gets frustrated or annoyed, is willing to work as long or as short as the user wants, and do this all basically free of charge,” Watson says.

Unlike parasocial relationships in which the celebrity has no idea the fan exists, AI chatbots “create an illusion of a two-way, reciprocal, connection,” Watson says. “AI bots do not have therapist eyes that look directly into a persons pain, joy and soul. AI bots also do not follow ethical standards of professional mental health organizations, and they are not mandated reporters when a serious crisis arises.”

While children are typically taught to delay gratification and manage their uncomfortable feelings for a period of time, as will be required by society when they’re older, AI bots do not provide any such boundaries, Watson says.

“The AI bot interaction can become prioritized over human interaction, as humans tend to be more confrontational,” Watson says. For those who suspect a loved one may be developing an unhealthy attachment to AI, Watson says warning signs include social isolation, a “blurring of reality,” and neglect of real-life responsibilities, much like with other addictions.

“They may spend an excessive amount of time using the AI bot, whereby they lose track of time, have eye or other physical strain, develop headaches, have disrupted sleep patterns, and may feel overall fatigued,” Watson says. “The AI user may also become anxious when away from the AI interaction, and they may have a sense of emptiness, irritability and or sadness when restricted from AI bot use.”

Parasocial relationships with AI tools aren’t just happening with the chatbots that have exploded onto the scene in the last few years, however. Human-AI interaction expert Julie Carpenter, PhD, a Research Fellow at the Ethics + Emerging Sciences Group, notes that these one-sided relationships “show up in many different scenarios: elder-care robots designed to soothe loneliness, in workplace AI systems that present themselves as ‘collaborative teammates,’ and in the digital reconstructions of the dead that quietly rewrite how someone is remembered.”

“These are all one-sided relationships, but theyre engineered and designed very purposefully to feel mutual,” she says. “Some people might assume theyre immune to this kind of thing, but the emotional incentives of genAI are designed intentionally to lean on familiar social cues — matching conversational timing, personality mirroring, even mimicking warmth and support — because those signals reliably boost user engagement and retention.”

The interactions feel personal, Carpenter says, because the AI systems use “familiar conversational rhythms: timely replies, small confirmations, and the kinds of back-and-forth markers people rely on in human conversation.” What concerns her is “the way companies treat those reactions as useful signals to track, predict, and monetize.”

“Emotional attachment becomes another vector for data collection, shaping how long someone stays engaged, how much they disclose, and how easily they can be steered. These systems arent just performing warmth, theyre also logging the moments when a person lingers, hesitates, or opens up, and folding that information back into engagement metrics. Parasocial bonds with technology arent a personal failing; theyre an expected outcome of systems built to optimize attention and retention,” she says.

“And our relationship to these systems wont stay static. As AI becomes a more common social actor, in homes, jobs, and intimate spaces, the cues we respond to will shift, and our expectations of interactions will shift with them,” Carpenter says. “The question isnt ‘Could this happen to me?’ but ‘What does it mean when everyday tools are designed to read us closely while giving the impression they understand us?'”