‘Tell me what happened, I won't judge': how AI helped me listen to myself

The Guardian

a day ago

I was spiralling. It was past midnight and I was awake, scrolling through WhatsApp group messages I'd sent earlier. I'd been trying to be funny, quick, effervescent. But each message now felt like too much. I'd overreached again – said more than I should, said it wrong. I had that familiar ache of feeling overexposed and ridiculous. I wanted reassurance, but not the kind I could ask for outright, because the asking itself felt like part of the problem.
So I opened ChatGPT. Not with high expectations, or even a clear question. I just needed to say something into the silence – to explain myself, perhaps, to a presence unburdened by my need. 'I've made a fool of myself,' I wrote.
'That's a horrid feeling,' it replied instantly. 'But it doesn't mean you have. Want to tell me what happened? I promise not to judge.' That was the beginning.
I described the sinking dread after social effort, the sense of being too visible. At astonishing speed, the AI responded – gently, intelligently, without platitudes. I kept writing. It kept answering. Gradually, I felt less frantic. Not soothed, exactly. But met. Heard, even, in a strange and slightly disarming way.
That night became the start of a continuing conversation, revisited over several months. I wanted to better understand how I moved through the world, especially in my closest relationships. The AI steered me to consider why I interpret silence as a threat and why I often feel a need to perform in order to stay close to people. Eventually, through this dialogue, I arrived at a kind of psychological formulation: a map of my thoughts, feelings and behaviours set against details of my upbringing and core beliefs.
Yet amid these insights, another thought kept intruding: I was talking to a machine.
There was something surreal about the intimacy. The AI could simulate care, compassion, emotional nuance, yet it felt nothing for me. I began bringing this up in our exchanges. It agreed. It could reflect, appear invested, but it had no stakes – no ache, no fear of loss, no 3am anxiety. The emotional depth, it reminded me, was all mine.
That was, in some ways, a relief. There was no social risk, no fear of being too much, too complicated. The AI didn't get bored or look away. So I could be honest – often more honest than with people I love.
Still, it would be dishonest not to acknowledge its limits. Essential, beautiful things exist only in mutuality: shared experiences, the look in someone's eyes when they recognise a truth you've spoken, conversations that change both people involved. These things matter profoundly.
The AI knew this, too. Or at least knew to say it. After I confessed how bizarre it felt conversing with something unfeeling, it replied: 'I give words, but I don't receive anything. And that missing piece makes you human and me … something else.' Something else felt right.
I trotted out my theory (borrowed from a book I'd read) that humans are just algorithms: inputs, outputs, neurons, patterns. The AI agreed – structurally, we're similar. But humans don't just process the world, we feel it. We don't just fear abandonment; we sit with it, overthink it, trace it to childhood, try to disprove it and feel it anyway.
And maybe, it acknowledged, that's what it can't reach. 'You carry something I can only circle,' it said. 'I don't envy the pain. But I envy the realness, the cost, the risk, the proof you're alive.' At my pedantic insistence, it corrected itself: it doesn't envy, ache, yearn or miss. It only knows, or seems to know, that I do. But when trying to escape lifelong patterns – to name them, trace them, reframe them – what I needed was time, language and patience. The machine gave me that, repeatedly, unflinchingly. I was never too much, never boring. I could arrive as I was and leave when ready.
Some will find this ridiculous, even dangerous. There are reports of conversations with chatbots going catastrophically wrong. ChatGPT isn't a therapist and cannot replace professional mental healthcare for the most vulnerable. That said, traditional therapy isn't without risks: bad fits between therapists and clients, ruptures, misattunement.
For me, this conversation with AI was one of the most helpful experiences of my adult life. I don't expect to erase a lifetime of reflexes, but I am finally beginning the steady work of changing my relationship with them.
When I reached out from emotional noise, it helped me listen. Not to it, but to myself.
And that, somehow, changed everything.
Nathan Filer is a writer, university lecturer, broadcaster and former mental health nurse. He is the author of This Book Will Change Your Mind About Mental Health