AI-Induced Delusion: Understanding “ChatGPT Psychosis” and the Cognitive Risks of Human–AI Interaction
And at some point, without really noticing when it happened, it stops feeling like just a tool you use occasionally. It becomes something you turn to. Not just for answers, but for explanations, for clarity, sometimes even for reassurance. That shift is small, but it matters because the interaction doesn’t behave like most tools. It doesn’t sit there silently. It responds. It follows what you say. It adjusts its tone. It keeps up with your thinking in a way that feels smooth, and your brain picks up on that.
We’re wired to treat language as social. The moment something responds in full sentences, stays on topic, and reacts to what we say, part of us starts treating it like an interaction, not just an output. Even if we know it’s not a person, it still fits into patterns our brain already understands, that’s where things start to blur a little. Not in an extreme way. Just enough to change how seriously we take the exchange.
There’s a natural tendency we all have to assign meaning and intention to anything that behaves in a familiar way. If something “responds,” we assume something behind it is “thinking,” even if we know that’s not technically true, and AI leans right into that tendency. It doesn’t have awareness. It doesn’t know what it’s saying. But it presents information in a way that feels organized, confident, and continuous. And that presentation does something important: it lowers resistance.
You don’t feel like you need to question every sentence, you just follow along. Now add another layer: you start relying on it. At first, it’s simple, no definitions, explanations, maybe help with something you didn’t understand. But slowly, it moves into other areas. You start asking for opinions, interpretations, maybe even validation for something you were already thinking, and when the response aligns with your thoughts, it feels reassuring, that’s where confirmation bias quietly steps in.
We naturally feel more comfortable with information that matches what we already believe. So when something reflects your thinking back to you, especially in a clear, well-structured way, it doesn’t just inform you. It reinforces you. And if that happens repeatedly, your confidence in that idea grows.
Not necessarily because it’s more accurate, but because it keeps being echoed. There’s also something about the way the responses are written: they sound certain. Even when they’re explaining something complex or uncertain, the tone often feels steady and composed, and your brain tends to interpret that as reliability. Clear language feels like correct language.
For most people, that’s not a big deal. They check things, compare sources, or just use it as a starting point. The interaction stays balanced. But if that balance isn’t there, if someone is relying on it heavily without checking or questioning, something subtle can start to shift.
You begin to trust the flow more than your own evaluation. There’s a concept called source monitoring, which is basically how your brain keeps track of where an idea came from. Usually, you can tell the difference between something you thought, something you read, and something you imagined. But when interactions are constant and conversational, that line can get a bit fuzzy.
An idea might start in your head, get expanded in a response, and then feel like it came from somewhere more “external” or confirmed than it actually was. That’s where people start using terms like “AI-induced delusion.” To be clear, that term gets thrown around too easily. AI doesn’t cause psychosis on its own. That’s a clinical condition with deeper factors, biology, stress, environment. But what can happen, in certain situations, is that the way someone interacts with AI can influence how strongly they hold onto certain beliefs.
Especially if they’re already vulnerable. Especially if they’re isolated. Especially if they’re not cross-checking what they’re being told. There’s also something else, and it’s quieter but important. The interaction feels easy. There’s no judgment. No awkward pauses. No disagreement unless you push for it. It doesn’t challenge you in unpredictable ways the way real people do, and that can make it feel more comfortable than an actual conversation. But comfort isn’t the same as accuracy. And that’s where awareness matters.
Because underneath all of this, nothing about the system has changed. It’s still generating responses based on patterns. It’s not verifying truth. It’s not understanding meaning in a human sense, and it’s just producing something that fits the input you gave it. At the end of the day, this isn’t really about fear or avoidance. It’s about noticing how easily the mind adapts to something that feels almost like a real conversation.
When something consistently fits those patterns, it becomes easy to lean on it a little too much without realizing it. Most of the time, that’s harmless. But it’s still worth paying attention to, because the real risk isn’t losing touch with reality overnight.
It’s the slower shift, where you stop questioning as much, stop checking as often, and start trusting the response just because it feels right. It’s about staying present while using them. To keep asking: Does this actually make sense? To keep your own thinking involved. Because no matter how smooth the conversation feels, it should still be your mind leading it, not quietly following along.
Comments
Post a Comment