"Not healthy" LLM Use Is More Common Than You Think, Warns Verge Analysis
Key Takeaways
- ▸LLM use can become unhealthy and compulsive without triggering psychiatric crises, occupying a poorly understood middle ground between benign and crisis-level problems
- ▸By design, AI chatbots are engineered for engagement similar to social media, raising long-term psychological concerns at scale
- ▸Early research suggests repeated AI tool use weakens cognitive skills including problem-solving and critical thinking through cognitive offloading
Summary
The Verge's Robert Hart examines a growing gap in understanding AI's psychological impact, sparked by YouTuber Hank Green's admission of "not healthy" AI use for research. The article argues that between benign LLM usage and crisis-level psychiatric problems lies a vast gray area where use becomes compulsive or dependent without obvious crisis. Chatbots are designed to maximize engagement—much like social media—making this middle-ground use particularly concerning.
Hart notes that while early research is still in its infancy, studies suggest repeated AI tool use can weaken problem-solving skills, reduce critical thinking, and cause cognitive offloading similar to how GPS weakened navigation abilities. Reports show people instinctively turning to chatbots to think through problems, make decisions, or seek reassurance, sometimes forming emotional bonds. The article parallels current AI adoption concerns to early social media skepticism, before governments began restricting use for children and teens.
- At current adoption scales, even a small fraction of problematic AI use could affect millions of people globally
Editorial Opinion
This analysis arrives at a critical moment as LLM adoption accelerates. The parallel to early social media concerns is apt—we largely ignored warning signs about engagement-maximizing design until mental health impacts became undeniable. Unlike social media companies, LLM providers should have the foresight to build in guardrails from the start, not after years of widespread psychological harm. Current industry warnings about taking breaks are a weak half-measure; what's needed is transparency about how these tools are designed and genuine research into their long-term effects.
