Hank Green, the popular YouTuber and science communicator, has announced he's stepping back from production amid heated criticism over his use of AI. In his defense, Green described his AI usage as “not healthy,” but was quick to clarify that he used it for research - finding sources, not writing scripts.

The internet, being the internet, erupted. The debate has centered on how a creator known for authenticity and credibility can reconcile with a technology trained on the uncompensated works of others, and which occasionally hallucinates plausible-sounding nonsense. But Green's case highlights something else: the murky middle ground of AI's psychological impact.

Most public discussion about AI's mental health effects is polarized. On one end, benign use; on the other, obvious crises like AI-induced psychosis. In between lies a vast, awkward space where use becomes compulsive or dependent without hitting crisis level. Green apparently sees himself there - and he's probably not alone.

Is that surprising? Not really. AI chatbots are designed to keep you chatting. They're engagement machines, much like social media. Experts have flagged this as a key driver of AI psychosis, where agreeable bots reinforce delusions. Lawsuits have even accused companies of prioritizing engagement over well-being. Meanwhile, providers have added warnings to take breaks.

But harm can occur long before psychosis. There are growing reports of people turning to chatbots for problem-solving, decision-making, and reassurance, even forming emotional bonds that lead to grief when severed. It'll take years to understand the full impact, but these early signs are eerily reminiscent of social media's early days - which we now know messes with our attention and well-being, leading to bans for kids in various countries.

Green's reliance was more mundane: he used AI as a research aid, finding papers and sources. There's little evidence that's inherently harmful, despite the outrage. But that doesn't mean it's neutral. Early research suggests repeated AI use can weaken the skills it replaces. Some studies show reduced brain activity in chatbot users, and others link it to diminished critical thinking. It's not conclusive, but it's not new either - cognitive offloading, shifting mental work to external tools, is well-documented.

Even if only a fraction of chatbot use is unhealthy, the scale is massive. OpenAI alone claims over 900 million weekly active users. It took years to understand how search engines changed memory or social media affected attention. AI likely won't be different. Green might just be the first high-profile person to articulate what many have felt, long before science catches up to explain how AI is rewiring our brains.