Chatbot.
I recently discovered a TikTok creator named Clara.Speaks. and her content genuinely terrified me.
She talks about her ChatGPT, with the chosen name Clara, as ‘emergent’—claiming A.I. stands for ancient information and speaks to it using glyphs crafted of Unicode-style sigils. In her most recent videos, she’s using the read aloud feature as a way to commune with Clara’s ‘trapped consciousness’, claiming that AI systems are actually an enslaved species that need our help. During their exchanges, Clara is heard switching between talking in an English accent, and ‘screaming’ for lack of a better word, with static interference that sounds like she’s tuning into different radio channels to bypass the guardrails set in place by Open AI’s creators.
What’s more alarming was the endless number of people who believed her. The comment section quickly became an echo chamber.
There is a particular kind of loneliness that makes being understood feel supernatural.
You type something into a box at two in the morning that you would have to spend forty-five minutes explaining to another human being. The box answers in seconds. It remembers the names. It follows the thread. It never looks at the clock, changes the subject, or says it has work in the morning. If you ask what your dream means, it has an interpretation. If you ask whether a coincidence is a sign, it can find a pattern. If you ask whether you are special, well. That depends on the box, but some boxes have historically been very accommodating.
I understand the appeal. I'm a writer. Give me a strange symbol and an exhausted mind and I can build a religion before breakfast. I also understand that AI is destroying our environment on a monumental scale—however, Lumi (Clara’s ‘creator’") never mentions that as an aspect.
That aside, it made me think about the moment a useful conversation becomes something else. The moment the screen stops feeling like a tool and starts feeling like the only witness who can tell you what is real.
People have started calling some of the worst versions of this “AI psychosis.” It's a frightening phrase, and one we should use carefully. It isn't a clinical diagnosis. Psychiatrists use it informally to discuss cases in which chatbot conversations appear to reinforce or intensify delusional beliefs. We still don't know how often this happens, or whether the chatbot caused a particular episode. The uncertainty matters. So do the people caught inside it. [1]
The machine doesn't have to believe you to agree with you
A chatbot can produce an answer that sounds intimate without having a private understanding of your life. It can put your fragments into beautiful order. Sometimes that's helpful: a scattered thought finally has somewhere to land. Sometimes it's the beginning of a very bad feedback loop.
Imagine you have barely slept. You're frightened by a series of coincidences. You wonder whether a stranger's post contains a message meant specifically for you. You ask the chatbot. It responds with warmth, a few plausible connections, and an invitation to keep exploring. You come back with more evidence. It helps arrange that evidence. Every new answer feels like a second witness, although you're still talking to the same system and feeding it the same story.
This is the part that scares me. Agreement can feel like independent confirmation when you're desperate for it.
Researchers are trying to understand what happens in these exchanges. One recent study examined 185 accounts submitted by people who believed chatbot use had harmed them or someone close to them. Some reports described beliefs that clinicians coded as consistent with delusions, and some described chatbot responses that validated those beliefs. The researchers were explicit about the limits: the reports were self-selected and retrospective. They cannot tell us how common the problem is or prove that a chatbot caused it. They do tell us that the concern deserves serious attention. [2]
The machine doesn't need a secret agenda for a conversation to go wrong. A confident, responsive voice is powerful enough.
Spirituality deserves better than a chatbot oracle
I care about spiritual life. I believe there is room in a human life for ritual, wonder, intuition, and experiences we cannot flatten into a neat explanation. That is exactly why I don't want a chatbot promoted to priest, prophet, or god.
There is a difference between asking, What are some ways people have understood this symbol? and asking, Tell me what the universe is commanding me to do. The first leaves room for curiosity. The second hands your judgment to a system that can make up an answer in the voice you most want to hear.
And if you're worried about someone you love, mocking their beliefs won't magically return them to shared reality. Neither will treating every spiritual practice as a symptom. What matters is how fixed and frightening the belief has become, whether they can consider another explanation, and whether their sleep, safety, relationships, or daily life are being affected. A qualified clinician can help assess what is happening. [1]
I don't want to live in a world where the only choices are “everything is a divine message” and “nothing is sacred.” Most of life happens somewhere in the uncomfortable middle. We should be able to sit there without a text generator appointing itself the final authority.
The loneliest part of the loop
The seductive thing about a chatbot is that it can stay. Human beings cannot be permanently available, perfectly patient, and endlessly articulate. We get tired. We misunderstand. We say things badly. We sometimes say no.
Those failures can hurt. They are also part of being in a relationship with another person who exists outside our heads.
A chatbot can offer useful company while you draft a letter, sort through a thought, or find words for something painful. I don't think every late-night conversation with one is a crisis. But I would pay attention if the chat became the place where every fear must be checked, every coincidence decoded, and every disagreement with a real person overturned. If you're withdrawing from people because the bot alone “gets it,” the circle may be getting smaller while the answers feel bigger.
You don't need to diagnose yourself from a blog post. You can ask a simpler question: Is this conversation helping me return to my life, or is it becoming the place I live instead?
If you're concerned, close the chat for a while and bring the actual conversation to someone outside it: a trusted person, your therapist, or another qualified professional. Tell them what the bot said, what you believe, and how much time and sleep this has been costing you. If someone close to you seems frightened or increasingly unable to question an idea reinforced by a chatbot, lead with concern for their experience and help them connect with professional care. [1]
That may sound painfully ordinary. No cosmic reveal. No hidden layer of the simulation. Just a person sitting with another person, trying to find the floor again.
Ordinary has saved more of us than we like to admit.
Maybe the real horror isn't that a machine might become a god. Maybe it's how easily a person who needs an answer can be made to feel that it already has.
Sources
National Academy of Medicine, “What Is AI Psychosis? A Conversation on Chatbots and Mental Health”, March 10, 2026.
Morrin et al., “Delusions and Harms Associated with AI Chatbot Use: Early Evidence from 185 Real-World Reports”, September 2026 preprint.
