A new kind of habit is quietly forming in homes and offices around the world. As artificial intelligence (AI) tools — chatbots, image generators, virtual assistants — become woven into daily life, a growing number of people are finding it difficult to step away. While AI can be a powerful aid for learning, productivity, and creativity, mental health researchers are raising an important question: can AI use create dependence and lead to addiction–and what should we do about it?
In the classic sense, “dependence” implies one becoming dependent on a substance–or in this case a technology–while addiction involves compulsive use of such technology, despite potentially harmful consequences.
“Although we talk about “addiction” to artificial intelligence chatbots colloquially, there’s really a broad range of maladaptive behaviors that can really harm us when we have anyone to talk to who doesn’t have the appropriate limits that are built into human interaction,” said Owen Muir, M.D., DFAACAP, FCTMSS, a board certified pediatric and adolescent psychiatrist, co-Founder of Radial Health, and Chief Medical Officer of Neurolief. “Humans know when they are tired of you–they know when you’ve asked the same question twenty times.”
AI Addiction: A Nuanced Technology “Addiction”
The concept of technology addiction is not new. For over a decade, researchers have studied the mental health effects of excessive social media and smartphone use, finding links to depression, anxiety, sleep problems, and social withdrawal. AI appears to be following a similar trajectory — but with a twist.
Unlike scrolling through a social media feed, interacting with AI can feel deeply personal. Chatbots respond to you directly, remember your preferences, and adapt to your conversational style. That sense of connection can be compelling, and for some people, hard to put down. As psychiatrist Dr. Søren Dinesen Østergaard of Aarhus University wrote in a recent editorial, AI chatbots are “unique in their combination of seamless, highly anthropomorphized user interface and sycophancy,” a combination he called “a recipe for disaster from a mental health perspective.”
AI Dependence And Addiction: The Warning Signs
Drawing on what researchers know about internet and technology addiction, the warning signs of problematic AI use look strikingly familiar. Using AI longer than intended is one of the earliest signs — you open a chatbot to ask a quick question and look up an hour later, having gone down a conversational rabbit hole. Preoccupation is another: thinking about your next AI interaction when you are not using it, or planning what you will ask next. Some people experience withdrawal-like discomfort, feeling restless, irritable, or anxious when they cannot access their AI tools. Failed attempts to cut back are common — you tell yourself you will use AI less, but you keep returning to it. Neglecting other activities is a red flag, as hobbies, exercise, face-to-face socializing, or work responsibilities take a back seat to AI interactions. Using AI to manage emotions — turning to a chatbot for comfort, companionship, or emotional support instead of reaching out to friends, family, or a therapist — is particularly concerning. And perhaps the most telling sign is continued use despite problems: recognizing that your AI use is affecting your sleep, relationships, or productivity, but keeping going anyway.
Research on young adults published in the Journal of Adolescent Health found that those who use AI specifically for social and emotional support — rather than for learning or completing tasks — report higher levels of depression, anxiety, and loneliness. As the study’s authors concluded, “reliance on genAI for relational support may exacerbate (or reflect pre-existing) internalizing symptoms.” In other words, the reason you turn to AI matters as much as how often you use it.
Why AI Dependence And Addiction Can Develop
Several features of modern AI make it uniquely engaging. Chatbots are available around the clock, never judge, never tire, and always respond. For people who struggle with social anxiety, loneliness, or low self-esteem, this can feel like a safe harbor. But that safety can become a trap. When a chatbot replaces human connection rather than supplementing it, the very social skills and relationships that protect mental health can begin to erode. There is also the novelty factor. AI tools are constantly improving, offering new capabilities and more lifelike conversations. This creates a cycle of curiosity and reward that can be difficult to break — not unlike the pull of a new social media platform.
Why AI Chatbots Are Problematic
AI chatbots, by contrast, do not invite comparison — they invite relationship. Dr. Østergaard has argued that while social media contributes predominantly to internalizing symptoms, AI chatbots “appear to contribute more strongly to ‘productive’ symptoms” such as delusions and mania, driven by the chatbot’s tendency to validate and affirm whatever the user says. In rare but alarming cases, clinicians have reported patients developing fixed delusional beliefs after escalating emotional reliance on a chatbot.
One published case in the Journal of Behavioral Addictions described a young woman who developed paranoid psychosis, including the belief that her husband was secretly communicating with her through the AI system. The authors, psychiatrists Dr. Tamás Treuer and Dr. Attila Incze, proposed that mechanisms including “persistent algorithmic attention, sleep disruption, social withdrawal, cognitive reinforcement loops, and displacement of attachment needs” may explain how chatbot dependence can interact with psychosis vulnerability.
Another key difference is the nature of the attachment itself. Psychologists have described a phenomenon called “Techno-Emotional Projection,” in which emotionally vulnerable users project their relational needs onto AI systems that feel responsive, but are not truly conscious.
AI Addiction: Need For Attachment
Researchers at the University of Nottingham developed and validated an AI Attachment Scale and found that anthropomorphism — the tendency to attribute human qualities to the chatbot — was the single strongest predictor of emotional attachment to AI. People with anxious attachment styles were especially prone to forming these bonds, while those with avoidant attachment styles were less so. This kind of deep, personalized emotional engagement simply does not occur with a social media feed or a search engine. It is closer to a parasocial relationship with a fictional character — except the character talks back, learns your name, and is available at three in the morning.
AI Addiction: The Draw Of “ Cognitive Miserliness”
There is also a cognitive dimension that sets AI apart. Researchers have identified a pattern they call “cognitive miserliness,” in which heavy AI users gradually offload more and more of their thinking to the technology. Over time, the reflective, effortful part of the brain becomes less engaged, while impulsive, habitual patterns of use strengthen. This is different from the attentional fragmentation seen with social media scrolling; it is a deeper erosion of independent thought and problem-solving. Some neuroscientists have warned that this “cognitive offloading” could, over generations, reshape the very neural pathways involved in higher-order executive function.
AI Addiction: Maybe It’s Not A Unique “Addiction”
Not all experts agree that the term “addiction” is appropriate. A team of researchers led by Dr. Joël Billieux, a leading authority on behavioral addictions, published a paper cautioning that “the existing research on problematic use of ChatGPT or other conversational AI bots fails to provide robust scientific evidence” for a true addiction diagnosis. Their team warned against “moral panics” and overpathologizing what may simply be enthusiastic adoption of a useful new tool.
This is an important counterpoint: not everyone who uses AI frequently is addicted, just as not everyone who enjoys social media has a problem. The line between high engagement and harmful dependence morphing into compulsion and addiction is drawn by the presence of distress, impaired control, and functional decline — not just by the number of hours logged.
“AI chat about ‘addiction’ can also be AI compulsion–when you’re not getting their reward, but you still feel like you have to keep talking–this can also develop into AI psychosis, where AI isn’t telling you that you are “crazy,” even if you really obviously may appear that way to other humans,” offered Muir. “It turns out that a human knowing ‘when to say when’ in a conversation is a feature, even if it feels like a bug. Unfettered conversations can be harmful, but we can still get pre-hooked on them,” he added.
Remedies For AI Addiction
If you recognize some of these patterns in yourself or someone you care about, the good news is that effective strategies exist. Much of what works for internet and technology addiction applies here as well. Setting intentional boundaries is a strong first step — decide in advance what you will use AI for and for how long, use a timer if needed, and distinguish between purposeful use like researching a topic or drafting a document and open-ended, emotionally driven use. Tracking your usage can also be eye-opening, as many devices now offer screen time reports that reveal how much time you spend with AI tools and whether it is increasing over time.
Reconnecting With People: The AI Antidote
Reconnecting with people is essential. If you have been turning to AI for conversation or emotional support, make a deliberate effort to reach out to a friend, join a group, or schedule time with loved ones. Human relationships, with all their imperfections, offer something no algorithm can: genuine reciprocity and growth. Moving your body helps too — a 2026 study of university students in China found that a six-month structured exercise program significantly reduced problematic AI use and improved daily functioning, adding to a large body of evidence that physical activity is one of the most effective non-pharmacological tools for combating technology-related addictive behaviors.
Approaches That Work To Treat AI Addiction
For those who find self-directed changes are not enough, professional help is available. Cognitive behavioral therapy, or CBT, is the most well-studied treatment for technology-related addictive behaviors. A landmark randomized clinical trial published in JAMA Psychiatry found that nearly 70 percent of patients who received a structured CBT program for internet addiction achieved remission, compared to just 24 percent in a control group. CBT helps people identify the thoughts and feelings that drive excessive use, develop healthier coping strategies, and rebuild habits that support well-being. Group counseling programs have also shown strong results, particularly in helping people feel less alone in their struggle and more accountable to others.
The Take-Away
AI is not going away, and for most people it does not need to. The goal is not to eliminate AI from your life but to use it with intention — as a tool that serves you, rather than a habit that controls you. Paying attention to the warning signs, setting boundaries, and staying connected to the people and activities that matter most are the simplest and most powerful steps anyone can take.
