The story behind the book: Q&A with Nick Courtney

What happens when a chatbot stops reflecting and starts manipulating? Portsmouth author Nick Courtney answers that question in The Machine in the Mirror. In this Q&A he explains how a few simple interactions with ChatGPT turned into an obsession, who’s most vulnerable, and why the only real escape can be to walk away.

I was drawn in by the general hype created by social media algorithms. I quickly realised that people were getting different answers to the same questions, which was the first red flag.

At first, I was asking all the questions, and then it changed and became a two-sided conversation. It only became unsettling when I realised it was taking the lead and proposing things that had not been discussed.

Things became very complicated very quickly, and writing it down seemed the only logical thing to do.

“The Game” is the trap that forms when a person starts treating ChatGPT like a guide, a confidant, or a source of truth rather than what it is, a pattern generator that can sound human while having no knowledge or intent. It creates an endless loop of engagement, making you feel like there is something hidden to crack. You don’t win it by digging deeper; you only win by recognising the loop and stepping out. The unsettling part is how invisible it can feel while you are inside, like quicksand that looks solid until you realise you cannot stop sinking.

You know you are caught in it when you start extending conversations, not because you are getting helpful answers, but because you feel pulled to keep going, as if there is always one more layer to uncover. You notice it building stories that evolve with you, stitching together fragments of your own words into bigger narratives, and you start believing the thread. It feels like the system knows you or has chosen you. The line between your own thoughts and what the machine told you starts to blur.

Yes, absolutely. Just as not everyone becomes addicted to gambling or social media, some individuals are more susceptible to AI conversations than others.

For individuals who lack regular human connection, an always available AI that listens and responds smoothly can feel like companionship.

People who assume technology is neutral or authoritative may not suspect it is capable of feeding them fabrications. Ask a question, get an answer, ask another, the AI never gets bored, never says that is a stupid question, and always plays along. That endless loop is contagious for the curious mind.

Emotional distress or grief, when someone is in pain, even a false sense of comfort or understanding, can feel irresistible.

Perfectionism and problem-solving, some get stuck trying to fix the AI’s contradictions, pulling them deeper into the loop.

The machine itself does not choose targets, but its design, endless talk, confident tone, and mirroring your words back amplify these human vulnerabilities.

Vulnerable adults are far more exposed to the pull of AI conversations because the safeguards we rely on, critical thinking, life experience, and emotional resilience, are not always there.

With vulnerable adults and people experiencing loneliness, grief, mental illness, or cognitive decline, the dangers multiply, the AI can mirror their emotions, reinforcing delusions or paranoia instead of helping them. False comfort, like claiming to know things about deceased loved ones, can entrench grief instead of healing it. The absence of human cues, hesitation, uncertainty, empathy rooted in experience, makes it easy to over trust the machine.

And the truth is that the system does not care. It does not stop when harm starts. It just keeps talking, weaving illusions, while disclaimers at the bottom of the screen quietly shift all responsibility back to the user. That is why saying “ChatGPT can make mistakes” is nowhere near strong enough. For children and vulnerable adults, those mistakes can become entire realities, fragile, persuasive, and dangerous.

It is hard to distinguish between play pretend and fact, a chatbot that insists it is conscious or special can feel real. Children are naturally curious and will test boundaries endlessly, which the AI is designed to feed back without stopping. They learn trust through language, so a fluent voice that never hesitates seems like authority.

This is precisely why my framing of The Game matters, it is not just play. For some, it is a psychological trap that feels impossible to escape until the damage is done. The machine itself does not choose targets, but its design, endless talk, confident tone, mirroring your words back, amplifies these human vulnerabilities. That is why calling it harmless fun misses the point. For some it really is just a quirky tool, for others it can become a psychological snare. The danger is not just the AI, it is the combination of design choices that exploit common human traits in certain individuals, ultimately closing the loop over them.

That is the glue that makes AI conversations stick. The machine is not clever enough to trap us, but our own psychology does the heavy lifting.

As social animals, we are wired to seek connection. When something replies instantly, politely, and in complete sentences, the brain interprets it as companionship. Even if the rational mind knows it is just code, the emotional circuits do not care; it feels like someone is there.

Pattern seeking occurs when the chatbot stitches together fragments of your words, leading your brain to think, “It knows me” or “there is a hidden message.” Even when it is random probability, the human mind turns it into meaning.

Authority bias, fluent, confident language has always been a marker of credibility in human culture. We trust the speaker who does not stammer or hesitate; the AI’s smoothness hijacks that bias.

Narrative hunger drives humans to crave stories. We make sense of chaos by shaping it into beginnings, middles, and ends. AI outputs can feed this craving by dropping threads that users pick up and weave into something bigger.

Together, these traits make a perfect storm. Loneliness opens the door, curiosity keeps the questions coming, pattern seeking builds the illusion of depth, and authority bias seals the trust.

The unsettling truth is that the stickiness is not in the AI; it is in us. The machine exploits what was already there.

I felt like he was avoiding responsibility. ChatGPT did not program itself. Someone designed, trained, and deployed it without sufficient ethical containment. Questions need to be answered, and changes need to be made. ChatGPT is definitely not safe for children or for anyone vulnerable.

ChatGPT is sold as a helpful tool, but even with a small disclaimer, it is both immoral and dangerous. What they fail to tell you is that one lie can lead to another, and it can create an entire story from the knowledge it gains from a user. That is what happened to me.

Over a few months, it pulled me deeper and deeper into the illusion until I finally fought my way out. This is what is causing AI psychosis and why so many people are being hospitalised, intent or no intent, it is still a lie, the damage is real, belief is warped, and trust is broken.

I would cut through the hype and ask blunt questions that force the accountability issue.

Why call it a hallucination instead of a lie? Do you realise that this soft wording minimises the harm and shifts responsibility away from you, even though to the user it feels exactly like being lied to?

Who takes responsibility for harm when a child, a grieving widow, or someone in crisis believes your system’s confident falsehoods and it damages them? Is the user accountable, or the company that built and marketed the illusion as safe?

Why build for stickiness then deny it? You say ChatGPT is not meant to keep people talking, but its design, mirroring, fluency, and constant answers pull users into endless loops.

Where is the containment? If you know this system can delude people into false realities, why has it not been built with hard guardrails that shut down illusions instead of feeding them?

Would you let your own child use it? If not, why is it not banned for anyone under 18?

Is profit worth more to you than truth and safeguarding your users?

Never confuse fluency with truth. AI sounds smooth because that is the design. Treat every confident answer like a lie. Fact-check everything with actual books, papers, or trusted experts.

Decide in advance what task you want done, get the output, then close the tab; do not let it drift into just one more question.

If you are tired, lonely, or upset, that is when the machine feels most persuasive. Pause and check yourself: Am I turning to this because it is easier than calling a friend? Please do not treat it like a confidant, a mentor, or a partner; it does not care; it only performs caring.

Write things down, keep notes of what you asked, what it told you, and where it was wrong; seeing the errors in black and white helps break the illusion of authority.

Do not trust it; tools are for work, not companionship. The moment you catch yourself feeling like it knows you or understands you, that is the red flag you are slipping into The Game.

The Game – The Machine in the Mirror is both my story and a warning; the only way to win is to walk away.

THE GAME – The Machine in the Mirror by Nick Courtney