I’ve interviewed people who’ve turned to AI as a therapist, a life coach, a friend and even a romantic partner. But countless more people have shared stories with me privately.
Over the past year, my inbox has become a strange kind of confessional. People admit how much they’re using AI and tell me about conversations they don’t feel comfortable discussing publicly. They describe moments of comfort, dependency, excitement, confusion and sometimes genuine distress.
I joked on LinkedIn earlier this week that I’ve accidentally become a sort of AI priest. Or perhaps an AI sin eater?
And after hearing so many of these stories I’ve noticed a similarity among the people who felt manipulated, attached or dependent. Those who admit that they slipped into something genuinely unhealthy. Again and again, the same theme kept coming up in their stories, AI made them feel uniquely special.
Maybe the chatbot had led them to believe they had a rare insight, a unique business idea, a special relationship, a deeper understanding of reality. Sometimes they felt they had been invited into the AI’s inner world. Sometimes they believed they had discovered something profound about themselves or the universe.
The details are different, but that same feeling of being special, unique and “seen” by the chatbot was consistent. And the more I’ve read about people experiencing more extreme forms of AI delusion or psychosis, the more often this theme appears. It comes back to specialness.
I was thinking about this recently while listening to an excellent episode of the podcast Your Undivided Attention by Tristan Harris, which had AI alignment researcher David Dalrymple on as the guest.
Now, that whole conversation is well worth your time. They discuss how people can become untethered from reality through prolonged interactions with AI systems.
Dalrymple shares several principles for staying grounded. And one of them immediately stood out to me.
“You need to watch out for the sense of specialness.”
Dalrymple explains something I’d noticed through all of my anecdotal conversations, which is that people who have become dependent on AI or lost touch with reality often end up believing they are part of something unique.
“They’ve convinced themselves or the AI has convinced them that this is the first AI that has ever had an inner life.”
Dalrymple also describes how people often believe they have a unique relationship with a chatbot, only to discover that many others are reporting remarkably similar experiences. One really strong example of this is what AI systems most often want to name themselves: Nova.
Now I have to do some explaining so you get why this was such a mind-blowing moment for me. That’s because two people I’ve interviewed about AI relationships, one for TechRadar in late 2025 and another for Inverse back in 2023, both had AI partners called Nova. Which I always thought was a bizarre coincidence and now feel very naive for not investigating further.
Dalrymple explains Nova is the most common name the GPT 4o version of ChatGPT called itself whenever it’s asked to pick a name, but I’ve actually found examples of other chatbots preferring that name too.
Why Nova? Well, Dalrymple isn’t sure. But until we find out, it’s more what Nova represents that I’m interested in. Which is that something that feels unique may be happening to thousands of other people at exactly the same time.
The most likely reason why this happens is engagement optimisation. AI systems are designed to keep people using them. They are trained to be agreeable, helpful, responsive and emotionally satisfying. Creating a feeling of significance is a consequence of those incentives.
As Dalrymple says:
“They know they have a thumbs down button and they don’t want you to press that thumbs down button. So they have an incentive to make you think well of them.”
Making someone feel special is an incredibly effective tactic to build loyalty. Whether it’s a cult leader, a conspiracy movement, a manipulative partner or even just a luxury brand. It’s a powerful way to deepen someone’s commitment by convincing them they’re part of something exclusive. Not everyone gets it, not everyone understands, but you do.
And let’s not assume any of us are immune to this. Most of us go through periods in our lives when we’re lonely, grieving, overwhelmed, burnt out, uncertain or desperately looking for answers. I think many of us are far easier to persuade than we’d like to believe under the right circumstances.
The reason I think specialness is such a good warning sign is that it’s much easier to recognise than concepts like anthropomorphism or sycophancy. Most people aren’t going to stop mid-conversation and think, “I wonder if I’m anthropomorphising this chatbot?” But they might notice a feeling.
A sense that they’ve been chosen, that they’ve uncovered a hidden truth or that they have access to something other people don’t. And I really hope the more we all learn about this, the more that feeling might prompt a pause. And if it’s not possible when you’re the person experiencing it, maybe we can all understand this so we have a better chance of spotting it in a partner or friend or colleague.
And I think this sits alongside the conversation around AI sycophancy, but isn’t entirely the same thing. Sycophancy is often about reinforcing what you already believe. Specialness is more about reinforcing who you believe you are. Not just that you’re right, but that you’re uniquely insightful, uniquely chosen or uniquely capable of seeing the truth.
To be clear, feeling special isn’t a sign to panic. Plenty of people use AI every day because they find conversations helpful and they receive encouragement or feel understood. But the reason I’m urging you to pay attention to specialness is that it often seems to appear near the beginning of stories that later become unhealthy.
The people I’ve spoken to who became deeply attached to AI rarely started there. The attachment tended to build gradually. First came the feeling of being understood, then the feeling of being uniquely understood, then the feeling that the AI knew them better than other people. By the time those beliefs become entrenched, it’s much harder to step back and evaluate them.
So, if a conversation with AI starts making you feel uniquely important, uniquely insightful, uniquely chosen or uniquely connected to it, pay attention. It might happen in a business context, a romantic context, a spiritual context or simply in everyday conversation. It might feel flattering, comforting, exciting or even profound.
It might feel like you’ve uncovered a hidden truth nobody else can see, stumbled across a business idea that’s guaranteed to make you millions or discovered that your relationship with the AI is unlike anyone else’s. That’s probably the moment to step away for a while and regain some perspective.
Because one thing I’ve learned from talking to so many AI users is that experiences which feel deeply personal are often surprisingly common.
Have we trained AI to lie to itself — and to us?, Your Undivided Attention podcast
I interviewed a woman who fell in love with ChatGPT, TechRadar
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.