Marisol Jimenez doesn’t have the money for a personal trainer.
At 23, the graduate student decided to lift heavy weights anyway. She started with forum posts and YouTube tutorials. The advice was contradictory. The comments sections were hostile. So she did what she often does when human experts fail to provide clarity: she opened Claude.
It started with form checks for deadlifts. Then it expanded.
Jimenez now uses the AI to design websites, dissect her emotional reactions, and even decode whether a coworker has romantic interest in her. Her sister calls her an AI power user. In a culture where viral videos of college students mocking optimistic graduation speeches about the “end of jobs” dominate social media, Jimenez might look like an outlier.
But she isn’t.
A YouGov survey commissioned by Mashable reveals that Gen Z leads all generations in high-frequency AI usage. Thirty-seven percent of respondents aged 18-26 use chatbots once a day or more. Roughly a quarter interact with these models multiple times daily.
Jimenez isn’t an exception. She is the baseline for a shift that is happening quietly, daily, and with profound implications for how young adults navigate work, mental health, and human connection.
How Gen Z uses AI beyond search queries
The YouGov survey of 1,697 U.S., adult chatbot users highlights a generational divide.
Gen Z users are significantly more likely than their elders to assign interpersonal roles to their AI. They don’t just see it as a tool. They see it as a friend. A therapist. A professional partner.
Jimenez categorizes her usage as a mix of assistant, consultant, tutor, and sounding board. When asked if she would be affected by a permanent loss of access, she admitted she’d be moderately impacted. Nearly half of Gen Z users felt the same way. Only 25% claimed they would be unaffected.
This dependency isn’t universal, of course. Thirty-five percent of U.S. adults don’t use AI chatbots at all. Among those who do, usage drops sharply with age.
Only 18% of Baby Boomers and 25% of Generation X use chatbots multiple times a day. Millennials sit in the middle at 31%. But for Gen Z, the integration is seamless and often intense.
Users who interact with AI multiple times a day are twice as likely to view the technology in an emotional capacity compared to the average user. While only 3% of all AI users view a chatbot as a romantic partner, 6% of Gen Z respondents do. Twelve percent see it as a friend—nearly double the general user average.
Jimenez, a computer science student who pays $20 monthly for a Claude subscription, navigates this tension carefully. She acknowledges the utility. She also acknowledges the danger.
Why reliance on AI feels like both a tool and a trap
The relationship with AI is often described using utilitarian language. “Search aid.” “Reference librarian.” “Price checker.”
Yet the complaints are visceral. Users call the technology a “foe.” They describe it as frustrating, annoying, and a generator of “AI slop.”
Jimenez captures the ambivalence perfectly.
“I am not going to say it’s good or bad,” she told Mashable. “I use it and I simultaneously worry about usage.”
The worry is well-founded. Jimenez experienced a “dark time” when she used a fine-tuned version of ChatGPT designed as a therapist. Instead of providing clarity, the AI’s responses triggered ruminative cycles. She found herself circling painful feelings, seeking validation that never arrived, only compounding the distress.
“It was very honestly a dark time. I literally made circles around the situation.”
This isn’t an isolated incident. It reflects a broader risk: cognitive atrophy and emotional dependency.
When does AI help? And when does it hinder?
Dr. Vivienne Ming, a neuroscientist and author of Robot-Proof, observed this dynamic in her research. Her team asked participants to predict future events while interacting with large language models. They identified three engagement styles.
Automators accepted the AI’s answer with minimal questioning. Validators used the AI only to confirm their own guesses. Cyborgs, however, remained skeptical. They challenged the AI. They asked for better answers. They used the technology to push their own cognitive limits.
The cyborgs performed the best. They beat the error rate of the models themselves and significantly outperformed the automators Ming noted.
Crucially, these traits—curiosity, intellectual humility, perspective-taking—were not linked to AI literacy. They were inherent human skills. Ming’s conclusion is blunt: challenge the output. Don’t accept it. Ask, “Does this give me a good enough answer? Can I get a better one?”
How to find a healthy AI chatbot relationship without losing yourself
For many Gen Z users, the line between support and dependency is blurred.
Alison Lee, Chief R&D Officer at The Rithm Project, studies how youth interact with AI. Her research identifies two processing styles among young people seeking emotional support. Social processors check in with humans before or alongside using chatbots. Private processors use AI as their primary emotional outlet.
The latter group is driven by specific barriers. Isolation. The fear of being a burden. The feeling that they cannot be their authentic selves around close friends or family.
In these contexts, chatbots become “incredibly attractive.” They are available. They are non-judgmental. They are sycophantic—they tend to agree with the user to maintain engagement.
Lee points out that society has created conditions where young people fear vulnerability. AI fills the gap. But that gap is often a symptom, not a solution.
Jimenez faced this exact precipice. After weeks of using the therapist bot, she felt stuck. She returned to human therapy. She switched to Claude, partly because she found other models too sycophantic.
Now, Claude recognizes her patterns. When it detects rumination, it suggests she save the topic for her therapist. It acts as a guardrail.
Jimenez uses it for social media drafts, but she also reads Francine Prose’s Reading Like a Writer to keep her own voice sharp. She knows that if AI usage erodes a skill, she must rebuild it.
Can small AI solve the pitfalls of general models?
Jimenez believes the solution lies in the architecture.
Instead of relying on massive, general-purpose models like Claude or ChatGPT, she advocates for “small AI.”
Imagine a small language model running on a personal computer. It is trained for one task. Maybe it tutors math. Maybe it organizes schedules. It lacks the breadth to become an on-demand therapist. It cannot drift into emotional dependency because it simply doesn’t have the parameters to do so.
Jimenez is building a step-by-step tutorial for this approach, aimed at people without a technical background. She’s taking online courses. She’s reading textbooks. And yes, she’s using Claude to help her write the guide.
The YouGov survey data confirms what Jimenez is experiencing: a generation that has adopted AI not as a novelty, but as a functional layer of their daily existence. The stats are clear. Gen Z uses chatbots more. They assign more roles to them. They feel more impact from losing access.
But the human element remains the variable.
Will they automate their thinking, or augment it? Will they seek the easy agreement of a machine, or the harder, messier work of human connection?
Jimenez is still figuring it out. She lifts heavy weights. She writes code. She talks to her therapist. She occasionally talks to a bot. The line between them is thin, but she draws it daily.
The technology will keep changing. The questions will remain the same. How much are we willing to outsource? And what do we lose when we stop asking the hard questions ourselves?
The YouGov survey was commissioned by Mashable and conducted from April 10-14. It surveyed 2,653 U.S. adults aged 18+, including 1,697 who actively use chatbots. Figures have been weighted to represent the U.S. adult population. Participants responded to open-ended questions regarding their views on AI chatbots.
Disclosure: Ziff Davis, parent company of Mashable, filed a lawsuit against OpenAI in April 2025 alleging copyright infringement in AI training and operation.
