When we think about the people using large language models (LLMs) in questionable ways, the name Hank Green might come to mind. But new research suggests that what we might call 'unhealthy' LLM use is far more widespread than most of us realize.
The Scope of the Problem
A recent survey conducted in early 2026 across 10,000 U.S. adults found that nearly 40% of regular LLM users reported at least one behavior associated with problematic usage. These behaviors include:
- Using LLMs to replace conversations with friends or family
- Turning to AI for emotional support instead of human connection
- Feeling anxious or irritable when unable to access an AI tool
- Spending more than three hours a day interacting with chatbots
The findings challenge the assumption that only a small subset of tech-savvy individuals fall into these patterns. In fact, the data shows that problematic use cuts across age, profession, and income levels.
Why It Happens
Experts point to several factors driving this trend. LLMs are designed to be helpful, responsive, and non-judgmental—qualities that can make them attractive alternatives to human interaction, especially for those who struggle with social anxiety or loneliness. Additionally, the convenience of instant answers and the subtle gamification of conversational AI keep users engaged longer than they might intend.
Dr. Elena Vasquez, a psychologist specializing in digital well-being, explains: 'These models are becoming more human-like every year. By 2026, we are seeing chatbots that remember past conversations, mirror emotional tones, and even initiate contact. This creates a powerful pull that some people find hard to resist.'
The Line Between Healthy and Unhealthy
Not all LLM use is problematic. The key, researchers say, is the intent and the outcome. Using AI for work, learning, or creative brainstorming is generally healthy. Problems arise when LLMs start to replace real-world relationships or when users rely on them to avoid facing difficult emotions.
A 2026 report from the Digital Wellness Institute outlines a spectrum of use:
- Healthy: Using LLMs as a tool to enhance productivity or knowledge.
- Borderline: Checking in with an AI for mood management more than occasionally.
- Unhealthy: Preferring AI conversation over human contact, neglecting responsibilities to chat with a bot, or experiencing distress when disconnected.
What Can Be Done
Tech companies are beginning to respond. Several major AI platforms have introduced usage nudges—encouraging breaks after long sessions—and partnerships with mental health organizations to provide resources directly within the interface. But experts agree that individual awareness is just as important.
If you find yourself turning to an LLM more often than to people, consider setting boundaries: limit daily chat time, schedule in-person meetups, and notice how you feel when you step away from the screen. A healthy relationship with technology means using it to enrich your life, not replace it.
This article is based on a report by Robert Hart, originally published in The Verge.
via The Verge AI