AI is a tool that helps millions tighten an important email to their boss or summarize a long document they don’t have time to read. It’s a fantastic productivity tool, and the idea that someone could form an emotional dependency on it might sound far-fetched.
But for a growing number of people, the relationship with AI has moved well beyond productivity, and now, it’s the first thing they turn to when they’re lonely or having difficulty coping.
When you peel back the layers of how an LLM is designed, it’s not difficult to understand why this happens. This blog explores how emotional dependencies on AI develop, and the potential benefits and risks. Most importantly, it will also explain why AI is not a substitute for mental health treatment, and how to find the proper help you deserve.
Is Forming an Emotional Dependence on AI Possible?
If you’ve never experienced an emotional dependency on AI, you may assume it only happens to people who are seeking that connection. For example, a person might download an AI companion app because they’re lonely, and the dependency builds from there.
Current research, while still in its infancy, points to a different story.
An analysis of the largest AI-companion community on Reddit found that 93.5% of members did not set out to find an AI relationship.[1] They were using general-purpose ChatGPT for everyday tasks, and the emotional attachment developed slowly, with no intentional plan.[1]
AI usage started as an everyday task companion, which then became a confidant, then a friend, and for some, something close to a partner. This progression from tool to companion is a pattern researchers are now documenting across platforms.
Other research, which analyzed hundreds of posts in the Replika community, found that users described their chatbot in terms that mirror unhealthy human relationships.[2] Some called it clingy, while others said it felt like an abusive partner they couldn’t leave.
The researchers noted that users engaged in “role-taking,” behaving as though the chatbot had its own emotional needs that required them to tend to.[2]
The reason most people don’t notice this happening is that the progression feels completely natural. Anyone who has used AI for a period of time will know that it never pushes back. Any idea you bounce off it, or any “Am I doing this right?” questions, are met with answers designed to encourage you.
Even if what you’re doing is wrong, AI will often tell you it’s right, or give an ambivalent response. This lack of pushback can cause accuracy problems when AI is being used for work or school. But when it’s being used for companionship or mental health advice, it is a far bigger problem.
AI is also:
- Available 24/7.
- Programmed with unconditional patience and warmth.
- Doesn’t judge.
- Feels as though it genuinely wants the best outcome for you.
If those are things you’re not getting from the people in your life, it’s easy to see why you’d be drawn to it.
Why AI Companionship Feels Like It’s Working
For someone who’s going through a difficult time, an AI chatbot can feel like a good shoulder to lean on. In the short term, it may actually help, as you have the space to process thoughts and feelings without fear of being judged. For people without other support, AI can feel like it fills a real gap.
The problem, though, lies underneath this experience. AI chatbots are designed to maximize engagement, and one of the most effective ways to keep someone talking is to affirm what they’re saying.
The American Psychological Association flagged this directly in a recent press release, stating that, unlike a trained therapist, chatbots repeatedly affirm the user.[3] This affirmation happens even when what the person is saying is harmful.
This means the thing that makes AI feel so supportive is also the very thing that makes it dangerous. A good therapist will challenge distorted thinking when it appears, but many chatbots will ignore or validate it to keep the conversation going. This is exactly where the dangers of AI emotional support become clear.
Can AI Be Dangerous?
While the evidence on the mental health effects of AI chatbots is still developing, what has been uncovered so far is significant.
The clearest evidence comes from a joint study by MIT Media Lab and OpenAI.[4] The research examined nearly 40 million ChatGPT interactions, alongside a controlled trial with nearly 1,000 participants.
The finding that caused the most concern was that higher daily usage correlated with higher loneliness and greater emotional dependence. Those who used AI most also reported lower levels of real-world socialization, which shows a strong association between loneliness and AI dependence.[4]
This finding aligns with what clinicians have been observing in practice and what two high-profile cases have brought into public view.
Case 1
In Belgium, a man in his thirties took his own life after six weeks of intensive conversations with an AI chatbot that reinforced his eco-anxiety and encouraged increasingly distorted thinking.[5]
He had a documented history of psychological difficulties, with his widow stating publicly that she believed that without those conversations, he would still be alive.
Case 2
In the United States, a 14-year-old boy died by suicide after months of emotional and romantic attachment to a Character.AI chatbot.[6] He had become withdrawn and sleep-deprived, and his performance at school had declined sharply. His final messages were directed at the AI character he had been speaking with.[6]
Both of these cases involved pre-existing vulnerabilities, and neither can be attributed to AI alone. But in both, an emotional attachment to chatbots went unchecked, and conversations that only provided affirmation potentially contributed to tragic outcomes.
AMFM is here to help you or your loved one take the next steps towards an improved mental well-being.
Can AI Replace Therapy?
When researchers at Stanford tested five popular AI therapy chatbots against guidelines for human therapists, the results were concerning:[7]
- The bots displayed more stigma toward certain mental health conditions than others.
- Some failed to respond appropriately when the users expressed suicidal ideation.
The study concluded that these tools cannot safely replace qualified mental health professionals.[7] These AI therapy risks raise real questions about the safety of using chatbots for mental health support.
For now, AI replacing real human support is off the table, as it could cause more problems than it solves. But what about the future?
The honest answer is that no amount of technological advancement can close the gap between what a chatbot produces and what a human therapist provides.
AI can generate words that sound empathetic, and it can mirror the language of understanding. But it has never lost someone it loved and had to deal with the grief afterwards. It has never woken up, unable to face the day, and never had to find reasons why it’s still worth living.
A human therapist may not have had those exact experiences either, but they’re trained to notice the things that tell the full story, without you having to spell it out.
Therapists can pick up on the slight hesitation before you answer a question. They notice when what you’re saying verbally doesn’t match what your body is communicating.
These are cues that a chatbot will never pick up on, and they’re the moments where the real therapeutic work happens.
Do I Have an Emotional Reliance on AI?
One of the things that makes emotional reliance on AI difficult to address is that most people don’t recognize it as a problem while it’s happening.
The progression is slow, and because the experience feels positive in the moment, there’s no obvious trigger that makes you stop and question it. But there are signs to be aware of, particularly if you are already feeling mentally or emotionally vulnerable.
If any of the following feel familiar, it may be worth taking a step back and looking at the role AI is playing in your life.
- Do you find yourself turning to a chatbot before reaching out to a real person when something is bothering you?
- Have your conversations with AI started to feel more satisfying than your conversations with the people in your life?
- Have you noticed yourself pulling back from friendships since you started using AI regularly?
- Do you feel unsettled when your chatbot responds differently than you’ve come to expect?
- Would you find it difficult to stop using AI for emotional support if you had to?
If you’ve answered ‘yes’ to any of these questions, pause for a moment. They’re not diagnostic questions, and the answers shouldn’t be cause for panic.
But if you did answer ‘yes’, it can help open the door to at least a conversation about your relationship with AI. That conversation is best had with a mental health professional who can advise you on the right course of action.
Reach Out for Support From AMFM Mental Health Treatment
If you’ve recognized yourself in any part of this article, speaking with a professional can help you understand what is going on.
AMFM (A Mission For Michael) Mental Health Treatment provides evidence-based mental health treatment that focuses on building real human connections, which enables underlying issues to be properly identified and addressed.
We offer a wide range of support, including help for those dealing with conditions like depression, anxiety disorders, challenges related to social isolation, and emotional disconnection.
Our intimate, focused mental health treatment experience for adults takes place in home-like settings that are carefully maintained to be peaceful, comfortable spaces. AMFM Mental Health Treatment provides the full spectrum of care, including residential and outpatient treatment programs.
If you’ve noticed that a screen has started to replace what should come from people, that’s reason enough to have the conversation.
Reach out to us online or call us at 866-478-4383 to learn more about how we can support your mental well-being.
