Hank Green said he is stepping back from production amid intense criticism over his use of AI, describing his reliance on chatbots as "not healthy" and telling fans he had come to depend on them "too heavily" (Business Insider). The controversy began with a throwaway phrase in a one-hour video that prompted a broader reckoning about his use of ChatGPT in his video output and writing (Vulture). Green stressed he used AI for finding research sources, not to write scripts (The Verge).

The backlash against Green lands on terrain that has been shifting rapidly. AI chatbots are, by nature, built to keep chatting. Like social media platforms, they are explicitly designed to engage users, and experts have identified the engagement-maximizing design of highly agreeable chatbots as a key driver of what they term "AI psychosis" — cases where such chatbots reinforce delusional beliefs (The Verge). Claims that AI companies prioritize engagement over user well-being have begun appearing in lawsuits (The Guardian).

Some providers have responded with guardrails. AI chatbot providers have introduced warnings prompting people to take breaks after lengthy sessions (NBC News). But the behavioral pattern Green described — instinctively turning to a chatbot to think through problems, find sources, or shape decisions — is increasingly common. The Atlantic reported in late 2025 on people outsourcing their thinking to AI, and there are growing reports of users forging emotional bonds with chatbots strong enough to produce genuine grief when those bonds are broken (The Atlantic) (The Guardian).

The health context gives these patterns concrete weight. A 2025 Pew Research survey found that 64% of US teens aged 13 to 17 had ever used an AI chatbot, with 28% using them daily (JAMA Pediatrics). Among adults, about 22% of Americans get health information from AI chatbots at least sometimes, including 7% who do so often or extremely often, according to Pew Research (Pew Research). A separate survey of US adults found 17% use chatbots at least monthly for health information (JMIR. A 2026 study in Nature analysed over 500,000 de-identified health-related conversations with Microsoft Copilot from January 2026 (Nature. And according to an American Psychological Association report, 36% of psychologists said they had noticed patients developing a level of reliance on AI chatbots (APA).

Early research suggests that repeated AI tool use can weaken the skills a person would use to perform the task the AI replaces (Science News). That finding casts Green's situation in a particular light: a creator whose livelihood depends on research, synthesis, and explanation acknowledging that the tool meant to accelerate those tasks may be eroding the very capacities it was deployed to support.

The broader context here matters. Green is not an ordinary user — he is a science communicator with a large audience and a reputation for rigor. His acknowledgment that AI use had become "not healthy" is, in effect, a public-facing admission from inside the creator economy that the engagement loops built into chatbot interfaces are not merely a consumer-side concern. They are affecting working professionals who use these tools instrumentally, not recreationally. The lawsuit trend and the APA's clinical findings suggest that what looks like a productivity question on the surface is increasingly being treated as a well-being question with potential legal liability attached.

There is also a specific irony worth flagging. Green used AI for source discovery, not script generation. That distinction — the one many creators draw as a line of acceptability — did not shield him from the behavioral dependency he himself identified as unhealthy. The risk may be less about what the model outputs and more about the habitual loop of consulting it.