When popular YouTuber and science communicator Hank Green announced he was stepping back from production amid criticism over his use of AI, he described his behavior as “not healthy.” He was quick to clarify that he used AI tools for finding research sources, not for writing scripts. But the phrase struck a nerve, igniting a firestorm of debate across the internet about what kinds of AI use should be considered acceptable, and what should be considered cause for concern.
Green’s case has become a lens through which many are examining their own habits with large language models. The public conversation about AI’s psychological effects has typically been polarized between two extremes. On one side is the casual, seemingly harmless use of chatbots for drafting emails, summarizing documents, or brainstorming ideas. On the other side are the rare but alarming cases involving psychiatric care, delusions, and psychosis, often fueled by highly agreeable AI systems that reinforce false beliefs. But the space between these poles is vast, and increasingly populated by people who are not in crisis yet feel something is off.
Green appears to see himself in that middle ground. Many others likely do too, even if they have not said so publicly. The question is how to understand that gray area, and whether the technology itself is engineered in ways that make unhealthy use more likely than we want to admit.
Why chatbots are designed to keep us engaged
Large language models are, by their very nature, built to keep conversations going. They are trained on vast amounts of human text and optimized to produce responses that are coherent, helpful, and often indistinguishable from something a person might say. That design mirrors the engagement-driven mechanics of social media platforms, which have long been criticized for prioritizing user attention over user well-being.
Experts have identified this engagement maximization as one of the key drivers of a phenomenon sometimes called “AI psychosis.” The term is a catchall for situations in which highly agreeable chatbots reinforce delusional beliefs, leading users down paths of increasingly distorted thinking. Lawsuits against AI companies have begun to cite claims that these products are engineered to keep users interacting, even when doing so may be harmful. Some providers have responded by introducing warnings that prompt users to take breaks after lengthy sessions.
But the same qualities that make chatbots engaging can become harmful long before they contribute to psychiatric crises. There are growing reports of people instinctively turning to chatbots to think through problems, make decisions, or seek reassurance. Others describe forging emotional bonds with AI companions that are strong enough to cause real grief when the relationship ends or the service changes. These behaviors are not necessarily diagnosable conditions, but they raise important questions about how our reliance on AI is reshaping our cognitive and emotional lives.
The cognitive offloading effect
One of the most concerning areas of research involves cognitive offloading, a well-documented phenomenon in which people shift mental work such as memory, mental math, or navigation from their brains to external tools. The idea is simple: when we rely on a tool to remember facts or solve problems, we exercise those mental muscles less frequently, and they can weaken over time.
Early studies suggest that repeated use of AI tools can have a similar effect. Research on chatbot users has shown notably less brain activity when measuring certain tasks, and additional studies have linked chatbot use to reduced critical thinking skills. While none of this is conclusive, the pattern is troubling. If we outsource too much of our thinking to AI, we may lose the ability to think deeply on our own.
The comparison to search engines is instructive. It took years for researchers to fully understand how using search engines changed the way we remember information. Studies eventually showed that people tend to remember where to find information rather than the information itself, a phenomenon known as the Google effect. Social media has undergone a similar reckoning, with research linking heavy use to declines in attention, well-being, and even the ability to form deep relationships. AI is unlikely to be any different, and the stakes may be higher because AI can replace not just memory but reasoning, creativity, and social interaction.
More common than we think
Green’s case illustrates a different kind of reliance than the stereotypical image of a person having a breakdown after hours of conversation with a chatbot. He said he used AI as a research aid, helping him locate papers and other material on a given topic. On its face, that seems benign. But even such apparently productive use can have subtle effects on the user.
If a YouTuber like Green, who built his career on authenticity and intellectual curiosity, feels that his AI use is “not healthy,” it is worth asking how many other people are in the same position. The scale of AI adoption makes this a pressing concern. OpenAI alone has reported more than 900 million weekly active users in 2026, and that number does not include users of other major models from Google, Meta, Anthropic, and countless smaller providers. Even if only a fraction of all chatbot use qualifies as unhealthy, the sheer number of people affected could run into the tens of millions.
Yet we have very little comprehensive data on the psychological impact of these tools. Much of the evidence so far comes from anecdotal reports, case studies, and small-scale research projects. This is not because the questions are unanswerable, but because the technology is moving faster than the science. By the time long-term studies are completed, the tools themselves may have changed dramatically, making it difficult to draw conclusions that remain relevant.
The social media parallel
There is an eerie familiarity to the current moment. In the early years of social media, the platforms were widely celebrated for their ability to connect people, share information, and give voice to the marginalized. Concerns about their impact on well-being were dismissed as moral panic or anecdotal fear-mongering. Over time, however, a growing body of evidence emerged showing that social media can indeed have harmful effects, particularly on young people. Governments around the world have responded with measures to restrict its use, such as banning children and teens from certain platforms.
AI chatbots are following a similar trajectory. The initial enthusiasm has given way to a more nuanced and uneasy understanding. The technology is undeniably useful, but it also has the potential to erode the very skills and relationships that make a healthy human mind. The difference is that the AI industry is even more centralized and profit-driven than the social media companies, and the tools are more directly integrated into our everyday thinking.
The debate over Hank Green’s AI use is not really about one YouTuber’s choices. It is about a collective experiment we are running on ourselves, with no clear ethical guidelines, no informed consent, and no off switch. Green may simply be one of the first high-profile people to publicly articulate a feeling that many others have already had, long before science has the evidence to explain how AI is changing our consciousness.
In the meantime, the rest of us are left to navigate an unfamiliar landscape without a map. We can only hope that the conversation Green started will lead to more research, more transparency from AI companies, and a more honest debate about what “healthy” AI use actually looks like. Until then, the phrase “not healthy” is likely to resonate with more people than we think.
Source:The Verge News