Are We Already Addicted to AI Chatbots? One YouTuber's Confession Opens a Bigger Question

Key points

• Hank Green, a science YouTuber with millions of followers, said in 2024 that his reliance on AI chatbots felt "not healthy," though he used them only for finding research sources, not writing scripts.
• OpenAI reported more than 900 million weekly active users in 2025, meaning even a small percentage of problematic use affects hundreds of millions of people.
• Early studies suggest repeated AI tool use can weaken the mental skills the tool replaces, a pattern researchers call cognitive offloading.
• Public discussion of AI harm tends to cluster around obvious psychiatric crises, leaving a much larger middle ground of compulsive or dependent use largely unexamined.
• The pattern echoes the early years of social media, which took over a decade to attract serious scientific and government scrutiny.

Hank Green makes science videos for a living. Tens of millions of people watch him because they trust him to get things right. So when he recently admitted that his use of AI chatbots, the conversational software behind tools like ChatGPT, had become something he described as "not healthy," his audience noticed.

Green was clear about what he actually did: he used AI to help find research papers and source material, not to write his scripts. That is a fairly modest use. Yet his confession still sparked a furious online debate, partly because of what it signals about a creator whose brand rests on intellectual honesty.

Why does this matter beyond one YouTube channel?

Green's case points to a gap in how we talk about AI harm. Almost nobody sits in between. Right now the conversation splits into two camps: people who say chatbots are basically fine, and news stories about users who developed delusions or psychosis after months of intense chatbot contact.

Between those two poles is a large, less dramatic space where use becomes habitual, compulsive, or emotionally dependent without tipping into a psychiatric crisis. Green seems to be describing exactly that middle ground.

AI chatbots are, by design, built to keep you talking. Much like social media feeds that are engineered to hold your attention, chatbots are optimised for engagement. That quality is what makes them feel helpful. It is also what makes them sticky in ways users do not always notice until they try to stop.

Claims that AI companies put engagement ahead of user wellbeing have already appeared in lawsuits. Some providers now show warnings asking users to take a break after very long sessions.

What does the research actually say?

The honest answer is: not much yet, but the early signals are worth watching. Several studies suggest that regularly using an AI tool for a task gradually weakens the human skill the tool is replacing. This is a well-documented process researchers call cognitive offloading, where we shift mental work like remembering directions or doing mental arithmetic to an external tool. The worry is that with AI, the outsourced tasks are getting bigger and more cognitively central.

Other early work found measurably less brain activity in people using chatbots for certain tasks, and some studies link chatbot use to reduced critical thinking. None of this is conclusive. The research field is young.

But the shape of the story feels familiar. It took years for the scientific and policy world to take seriously the idea that social media was affecting attention, mood, and behaviour at scale. Governments are only now restricting children's access to those platforms.

OpenAI alone reported more than 900 million weekly active users in 2025. Even if only a small fraction of that group is developing unhealthy patterns, the raw numbers are enormous.

Green may simply be one of the first prominent people to put words to something many ordinary users have quietly felt, well before science has caught up with an explanation.

Common questions

Does using AI for research make you dependent on it?

Not necessarily, but early research suggests that any repeated outsourcing of mental work can gradually weaken the underlying skill. Using AI as a starting point rather than a final answer is a practical way to stay in the habit of thinking things through yourself.

How can I tell if my chatbot use is becoming unhealthy?

Researchers have not yet agreed on a clear threshold, but worth watching for are: reaching for a chatbot before trying to think through a problem yourself, feeling anxious when you cannot access one, or noticing that decisions feel harder without it.