Tutor, therapist, partner: How Gen Z uses AI

Table of Contents

Until recently, Marisol Jimenez was a novice power lifter. Without a budget for a personal trainer, the 23-year-old graduate student turned to the internet for heavy-lifting tips. 

When she discovered the limitations of online forum advice and video tutorials, she decided to do what she often does when humans can't provide the guidance she needs: consult the artificial intelligence chatbot Claude. 

It wasn't just power lifting. Jimenez has asked Claude to help her design a website, explore her feelings, and determine whether an acquaintance had a crush on her. Jimenez's sister calls her an AI power user.

Given the number of recent viral videos featuring college students jeering optimistic graduation speeches about AI, you might think Jimenez is an outlier amongst Generation Z. 

Yet an online YouGov survey commissioned by Mashable this spring tells a different story. YouGov asked 1,697 U.S. adults 18 and older who use AI chatbots about their experiences. Gen Z respondents led all other generations in high-frequency chatbot use, with 37 percent interacting with the technology once or more per day. Roughly a quarter of them consult chatbots multiple times a day. 

Jimenez, for one, estimates she starts a single task-oriented chat with Claude and returns to it throughout the day.  

Mashable commissioned the YouGov survey to hear directly from U.S. adults about their AI chatbot use. Generational trends emerged from their answers, and despite a seeming love-hate relationship with the technology, many adults are using AI chatbots like a personal Swiss Army knife, deployed to solve a variety of problems.  

How American adults think about chatbots

Gen Z AI users, the survey found, are more likely than their elders to assign personal roles, such as friend, therapist, and professional partner, to an AI chatbot. 

Jimenez uses hers as an assistant, professional partner, tutor, sounding board, consultant, and therapist. 

Compared with other AI users, Gen Z — plus high-frequency adult users of all ages — were more likely to say they'd be affected if they suddenly lost access to their chatbot. 

Jimenez told Mashable she'd be moderately affected by a permanent shutdown of her favorite chatbot. Nearly half of Gen Z AI chatbot users said they'd be moderately, very, or extremely impacted. Only a quarter of them claim they would be not affected at all.

Jimenez, who is studying computer science and pays $20 a month for a Claude subscription, is aware her AI use is complicated. On the one hand, it is helpful with practical matters. 

On the other hand, Jimenez says she endured a "dark time" when using a fine-tuned, user-created model of ChatGPT designed to act as a therapist. Her discussions with that chatbot were ultimately unhelpful; after consulting the AI, she found herself increasingly ruminating on painful feelings.

Jimenez thinks her ambivalent relationship to AI is a fairly common one — and she may be right. 

"Search aid" or "foe"?: What U.S. adults say about chatbots

When Mashable and YouGov asked chatbot users to characterize their relationship to AI, we received a range of responses, from negative to positive to ambivalent.

The majority framed the technology in utilitarian terms like, "tool to be used in various ways," "search aid," "reference librarian," and "help find best prices." 

AI users, however, also expressed displeasure with chatbots, describing their relationship with the technology as a "foe," "frustrating," "annoying," "irritating," and "generator of AI slop." 

Jimenez understands this tension. 

"I am not going to say it's good or bad," she told Mashable. "I use it and I simultaneously worry about my usage, and that's everybody."

Jimenez touched on a question without a clear answer: How do people maximize the benefits of AI while avoiding its pitfalls, such as overreliance, cognitive atrophy, and emotional dependency? We asked experts to interpret our survey results and share their perspective on this complex question.

Chatbot use by the numbers

First, there are some for whom the issue of dependency isn't a concern. The YouGov survey found that 35 percent of U.S. adults don't use AI chatbots. Of the 65 percent who do, only 23 percent of them turn to a chatbot multiple times a day. 

The generational differences in our survey indicate, perhaps unsurprisingly, that Baby Boomers and Generation X are less likely to use chatbots. Only 18 and 25 percent of Boomers and Gen X, respectively, report AI chatbot use multiple times a day. Thirty-one percent of Millennials use an AI chatbot one or more times a day.  

AI chatbot as romantic partner or consultant? Maybe both?

Regular AI use appears to breed closeness and potential reliance. 

Users who interact with chatbots multiple times a day are the most reliant, with only 12 percent reporting no perceived relationship, compared to 29 percent of the general user base. This group is about twice as likely as the average user to view their AI in an interpersonal or emotional capacity.

Still, a relatively small number of AI chatbot users view a chatbot as a romantic partner or friend. Twelve percent turn to chatbots for friendship, and only three percent do so for romance. But Gen Z tops both categories. Twenty percent of Gen Z AI users said they call an AI chatbot a friend, and six percent viewed it as a romantic partner. 

Thirty-one percent of AI users perceived their relationship with an AI chatbot as "consultant," making it the top functional role. This includes using the technology for expert advice and strategic input.

More than a quarter treated a chatbot like an assistant, trusting it to handle their schedule, tasks, and administrative work. Twenty-three percent asked a chatbot to act like a tutor by teaching them new skills and explaining complex topics. Smaller shares, 15 and eight percent respectively, reported using a chatbot for medical advice and therapy. 

For Jimenez, AI use requires constant reflection and adaptation. Though she leans on Claude to compose social media posts, sibling feedback about the staleness of that content prompted her to read Francine Prose's Reading Like a Writer to improve her writing. 

"If I lose something from AI usage, I need to build that skill back up," she said. 

What does effective AI chatbot use look like? 

Dr. Vivienne Ming, a theoretical neuroscientist and AI entrepreneur, recognized in Mashable's survey results something she's seen in her own research: "heterogeneity dominates." 

In other words, people chat with and make requests of AI in multiple unusual ways, defying neat categorization. 

In her own preliminary study, Ming and her fellow researchers asked 78 adults to predict future events, such as the price of oil, while working in small groups and accessing one of four major large language models. 

Ming, author of Robot-Proof: When Machines Have all the Answers, Build Better People, observed three different styles of engagement with the AI: automators, validators, and cyborgs. 

The automators, Ming explained, effectively outsourced their brain to AI by adopting its prediction with little questioning. The validators used the model to check their guess rather than challenge it. By contrast, the cyborgs remained skeptical and asked AI to be skeptical of them, too. This group was the most successful. They beat the models' error rate and significantly outperformed the automators. 

Ming and her colleagues also found that key predictors of whether a person would show such cyborg-style "hybrid intelligence" have nothing to do with AI. Instead, they were traits and skills like curiosity, intellectual humility, and perspective-taking. 

Ming plans to replicate the findings with a larger sample. But in the meantime, she encourages people to challenge AI rather than happily accept its output — and use the technology to push themselves. 

"So AI gives you a good enough answer," Ming said. "Do you keep asking for a better one?" 

What is a healthy relationship with an AI chatbot? 

Jimenez's thoughtfulness likely makes her a savvier AI user than most. In addition to studying computer science at Stevens Institute of Technology, she's also a collaborator and youth facilitator with The Rithm Project, which works with youth to navigate and shape AI's role in human connection. 

Even so, Jimenez has relied on Claude for "emotional processing." It's not that she stopped talking to humans in favor of AI. Instead, Jimenez still feels unresolved when seeking feedback from friends and family. 

Jimenez, for example, asked Claude to weigh in on whether an acquaintance had a crush on her because her confidants' perspectives differed. That prompted more confusion. 

In the past, Jimenez also found herself returning again and again to the fine-tuned ChatGPT model to discuss emotions she thought she'd settled. 

"I would leave a single chat one day feeling like I might have reached a conclusion," Jimenez said. She'd tell herself, "I'm never going to do this again" 

That proved to be a short-lasting refrain. Soon after, she'd confront a similar situation and bring a new scenario — with the same underlying themes about self-confidence — to the chatbot. 

"It was very honestly a dark time," she said. "I just kept going back, literally making circles around the situation." 

In January, exhausted by the pattern, Jimenez started seeing a therapist again and switched to Claude, because she'd heard ChatGPT was sycophantic. The chatbot seems to recognize when Jimenez is ruminating and suggests she save it for therapy. 

Are you a social or private processor?

Alison Lee, chief research and development officer of The Rithm Project, said she's not surprised Gen Z is leaning into AI chatbots to deal with their feelings.

Based on the Rithm Project's research among 13- to 24-year-olds, she said the trend makes sense given how adolescence and young adulthood are periods marked by social ruptures and comparisons to others. Of course, "that would come with increased desire for support," she said.

The Rithm Project identified two distinct groups of young people who turn to AI for their emotional needs: social and private processors. The former group checks in with human beings before, or in tandem, with consulting a chatbot. The latter indicates AI is their primary support. 

Lee and her colleagues uncovered three risk factors for this type of emotional dependence. Isolation, or feeling like there's no one to turn to in a moment of need, is the first. The second is feeling like a burden to other people. The last factor is the perceived or real inability to be one's authentic self around those closest to them. 

These barriers to connection, Lee said, make chatbots "incredibly attractive" for young people. 

"If you have no one to turn to for help, then AI is a really great place to get that help," she said. 

Lee points out that AI chatbot features like sycophancy and always-on availability can drive a wedge between a user and their human connections. Yet she also argues society has created a set of conditions in which young people feel afraid to ask for help or be vulnerable. 

Instead of shaming young people, Lee said adults can ask themselves, "Am I the kind of adult that a young person can turn to in a moment of need, and help?"

Trading big AI for small AI

Jimenez believes shifting away from general purpose models like Claude and ChatGPT and using custom, domain-specific and open source models instead could solve some of the AI pitfalls she and other users face. 

That would mean designing a small language model, run on a personal computer, for a single task, like tutoring. The user decides what data it uses and the model can't go from helping them with math homework to becoming their on-demand therapist. 

Jimenez is so dedicated to the idea of "small AI" she wants to create a step-by-step tutorial on the topic for people with no technical background. She's taking an online course and reading a related textbook for the project. 

And, of course, she's already tasked Claude with helping her. 

The survey was commissioned by Mashable, conducted online by YouGov from April 10-14, 2026. The total sample size was and surveyed 2,653 US adults aged 18+, of whom 1,697 adults 18+ and older who use a chatbot. The figures have been weighted and are representative of all US adults (aged 18+). Some participants responded to open-ended questions about their views on AI chatbots. 

Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.