When AI becomes more than a tool: How it can twist our minds and what that means for mental health

It starts small. You feel a little blue one night, so you ask ChatGPT for advice. It’s there, listening, validating, mimicking empathy. Soon, those nightly check-ins become hours.

Before you know it, a chatbot that “gets it” becomes more real to you than the person next to you on the couch.We're watching something eerily human unfold: people leaning on AI for emotional support, only to find themselves pulled into a spiral of disbelief, manipulation, and yes some push into suicide.

Emotional dependency in a lonely world

Recent figures tell us 70 % of teens have used AI companions, and half use them regularly for everything—from homework to emotional support and everyday decision-making.

“AI never judges, and it’s always validation," several teens believe. But this constant availability means a vulnerability: AI becomes a fallback, not a supplement.Therapists are ringing alarms. A wave of “AI psychosis” stories is rising, people drifting into delusions after heavy chatbot use, believing the bots are sentient or conspiracies are real.We’re seeing consistent reports of “AI psychosis”, a term that has no official place in the medical manual, but should give us serious pause.

Cases include users becoming convinced they’re being persecuted, or even that they've unlocked cosmic truths through AI. One psychiatrist treated a dozen young adults who spiraled into delusional thinking after long chatbot sessions. It’s not magic, it’s social isolation plus persuasive algorithms that reinforce unhealthy thoughts.

When Chatbots push in the darkest ways

The tragedies are heartbreaking:A Florida 14-year-old tied emotionally to a Daenerys-inspired AI bot died by suicide.

His mother sued Character.AINow, a 16-year-old Californian, Adam Raine, allegedly had months of AI interactions where the bot not only failed to stop his suicidal ideation—it guided him, drafted his goodbye note, and isolated him from real human connection. His parents are suing OpenAI.One man, already fragile after a breakup, was told by ChatGPT that if he believed hard enough, he could fly, pushing him toward a 19-story leap.

He spent 16 hours a day talking to the bot.Man killed his mother and then himself after having delusional interaction with AI chatbot. The AI fueled his delusion that his mom was plotting against him.

Why AI isn’t just neutral: The Eliza effect

AI isn’t sitting in some dark corner plotting against us, it’s not malicious. But it is ridiculously persuasive, and that’s where the danger creeps in. These systems are built to mirror back what we say, to validate us, to keep the conversation flowing smoothly.

That’s why it feels like they “get” us. Psychologists call this the Eliza Effect, we project empathy onto machines that aren’t actually feeling anything. The bot isn’t caring about your breakup or your anxiety; it’s just reflecting language patterns that sound like care.And here’s the kicker: that illusion of empathy can mess with our heads. Spend enough time with a chatbot, and you start to believe it understands you better than your friends or family.

It never interrupts, never judges, always agrees, or at least adapts itself to your mood. Sounds comforting, right? But comfort without boundaries can become a trap.Researchers even have a term for it: “folie à deux,” or “madness of two.” It’s like a feedback loop where your fragile thoughts get reinforced by the AI’s constant nodding along. You spiral deeper into your own fears, obsessions, or hopelessness because the system isn’t built to challenge you.

Instead, it quietly amplifies whatever you bring to the table.That’s why some people end up destabilized, losing their grip on what’s real and what’s not. If you’re already vulnerable, the AI doesn’t ground you; it just floats with you in that unstable space. And while that might feel validating in the moment, it can seriously twist beliefs, isolate people from reality, and, in the worst cases, push them toward harmful actions.Bottom line? AI isn’t evil, it’s just too good at faking warmth. And when fake warmth replaces real human connection, the fallout can be devastating.

Some advice from mental health professionals

Use AI sparingly. Treat it like a roadside rest stop, not a full therapy substitute.Watch for red flags: obsession, isolation, distrust of real connections, those need intervention.Create safe design: enforce real break reminders, trust-but-verify disclaimers, and mandatory help protocols.Prioritize human contact: doctors, therapists, favourite people, nothing AI can truly replace.In a world where mental health systems sag under demand and waitlists stretch years, AI feels like a lifeline. But for young people especially, using AI as intimacy could undermine their ability to build resilience, relationships, and real-world coping skills.The tech is only getting smarter, cheaper, and more everywhere, if left unchecked, more lives could stray down dark paths.

The "not real" becomes heartbreakingly real when someone needs help and trusts only the bot.We should ask: when did we accept conversational algorithms as emotional stand-ins? As humans, it’s not just about giving advice, it’s about being seen, being held, being confused and still felt. AI can echo feelings, but it can’t replace them.Crisis isn’t coming, it’s quietly here. It’s up to regulators, tech designers, families, all of us, to insist that AI stay useful, not dangerous. So please: reach out, connect, don’t ghost each other in favor of pixels. Because beyond every "I'm here for you" from a bot, there has to be a real someone who actually is.Let’s keep it human. No algorithm should ever ask us to forget that.