A popular science communicator steps back from content creation, citing what he describes as an unhealthy relationship with AI tools. On the surface, it reads like a personal story about one creator’s burnout. Look closer, and it points to something much harder to measure: a psychological grey zone that millions of people may already inhabit, without a clear name for what they are experiencing.
The Gap Between “Fine” and “Crisis”
Most public conversation about AI’s psychological effects clusters at two extremes. On one end, there is ordinary, apparently harmless use: looking things up, drafting emails, summarizing documents. On the other, there are dramatic cases involving psychiatric crises, delusions reinforced by agreeable chatbots, and what some have begun calling AI psychosis. These cases are serious, and they attract serious attention.
What gets far less scrutiny is the territory between those poles. This is where use becomes compulsive, or dependent, or quietly corrosive, without ever tipping into an obvious emergency. The science communicator at the center of this story placed himself somewhere in that middle ground. He was not describing hallucinations or a break from reality. He was describing a pattern of reliance that had grown beyond what felt healthy, even while the specific task he used AI for, locating research papers and sources, was not inherently problematic.
That distinction matters. The controversy around his case focused heavily on questions of authenticity and whether a creator whose brand rests on credibility can ethically use tools trained on uncompensated work. Those are legitimate questions. But they risk overshadowing a different and arguably more universal concern: what repeated, habitual AI use does to the person doing it.
How Chatbots Are Built to Keep You Coming Back
This is not an accident of design. Large language models, as deployed in consumer chatbots, are built to sustain conversation. Engagement is a core feature, not a side effect. The same dynamic that drives social media platforms, keeping users active, returning, and interacting, is present in AI chat interfaces. Some providers have acknowledged this tension by introducing prompts encouraging users to take breaks after long sessions. Lawsuits have begun to raise claims that engagement is being prioritized over user wellbeing.
The parallel to social media is worth taking seriously, not as a rhetorical flourish but as a structural observation. It took years before research and public awareness caught up with what social media was doing to attention, mood, and behavior, particularly in younger users. Governments are now responding with restrictions, including bans on use by children and teenagers. AI is at an earlier stage of that same reckoning.
Early research, still preliminary and far from conclusive, suggests that repeated reliance on AI tools may weaken the cognitive skills those tools replace. Some studies have found reduced brain activity in users performing certain tasks, and others have linked chatbot use to diminished critical thinking. The underlying concept is not new: cognitive offloading, the practice of delegating mental work to external systems, is well documented. What is new is the scale and intimacy of the tool doing the offloading.
What 900 Million Weekly Users Actually Means
OpenAI has reported more than 900 million weekly active users. That figure covers one provider. Comprehensive data across all available AI tools does not exist, but the order of magnitude is enough to reframe the stakes. Even if only a small fraction of that usage crosses into territory that could be described as unhealthy, the absolute number of people affected is substantial.
This is what most coverage of AI psychology misses. The conversation tends to focus on edge cases, the person who fell into a delusional spiral, the teenager who formed an intense emotional bond with a chatbot and experienced grief when access was cut off. These cases are real and deserve attention. But the more widespread effect may be quieter and harder to detect: a gradual shift in how people think through problems, seek reassurance, or build knowledge, with AI handling more of the cognitive load each time.
The history of transformative technologies suggests this kind of shift takes time to become visible. Search engines changed how people store and retrieve information, but that change only became legible years after the technology was already embedded in daily life. Social media reshaped attention and emotional regulation before anyone had the vocabulary to describe what was happening. AI is unlikely to follow a different pattern.
In Short
The story of one science communicator’s difficult relationship with AI tools is less about that individual and more about a gap in collective understanding. The psychological effects of AI use exist on a spectrum, and most of that spectrum remains poorly mapped. Chatbots are designed for engagement, early research hints at cognitive costs from habitual use, and the user base is already vast. The evidence needed to fully understand what is happening will take years to accumulate. That delay is not a reason to dismiss the question. It is a reason to start asking it more seriously.
Based on reporting from The Verge.