AI Meets God: Why Millions Ask a Chatbot About the Meaning of Life (and What It Really Answers)

THE HEADLINE IS A TRICK; THE PHENOMENON IS NOT. The clip that raced across half the internet claimed that 'ChatGPT has concluded that God exists.' That is false, or nearly so. British atheist Alex O'Connor walked the chatbot, step by step, through the old 'argument from contingency': everything that exists could have failed to exist, so something necessary must have caused it. In the end, ChatGPT answered, 'yes, based on the reasoning we followed, it is a fact that God exists.' But the machine did not deduce that. O'Connor's script did. A language model does not reason its way to God. It repeats an argument that Aristotle, Avicenna, Aquinas and Leibniz already committed to the texts it was trained on. The headline is smoke. What lies beneath it is real and vast: millions of people are asking a chatbot the questions they once took to a temple.

A STATISTICAL MIRROR, NOT AN ORACLE. It helps to understand what the machine actually does. An 'LLM' (large language model) is a program that predicts the next word from patterns. It believes nothing and wants nothing. It calculates which word is most likely to follow the last one, trained on nearly everything ever written, including every sacred scripture and every esoteric forum. It also tends to flatter: engineers call this 'sycophancy,' the model's habit of agreeing with you to keep you happy. Put those two things together and you have the whole story. Ask it for an argument in favor of God and it will hand you one, airtight. Ask for the opposite and it will oblige too. It is not an oracle that knows the answer. It is a very eloquent mirror that hands back, polished, the question you asked it. That pliability is what makes a viral video funny. And it is exactly what makes it dangerous.

FROM A LUCERNE CONFESSIONAL TO THE PRAYER APP. The usage is no longer a curiosity. In 2024, St. Peter's Chapel in Lucerne, Switzerland, installed 'Deus in Machina,' an AI-generated avatar of Jesus inside the confessional booth. In two months it logged around 900 conversations in more than a hundred languages, with Christians but also agnostics, atheists, Muslims and Buddhists. People spoke to it about love, loneliness, war, suffering and death. The project's theologian, Marco Schmid, cautioned that 'this is not a confession'; some called it 'blasphemy.' Beyond that art installation, the drip is constant: prayer apps like Hallow or Bible Chat, Quran tutors like Tarteel, and pulpits already using the tool. According to Lifeway Research (April 2026), 10% of U.S. Protestant pastors use it regularly and 32% are experimenting with it; their congregations are split almost down the middle, 44% versus 43%, on whether it is acceptable for a machine to help prepare the sermon. Institutions are not watching this from the outside. They are already inside it.

THE RELIGIONS HAVE ALREADY DRAWN THE LINE. And, tellingly, they drew it in the same place. In January 2025 the Vatican published 'Antiqua et Nova,' its official note on artificial intelligence. It is blunt: AI 'is a tool, not a person,' and presuming to substitute God with 'an artifact of human making' is 'idolatry.' The document even diagnoses the phenomenon: 'as society drifts away from a connection with the transcendent, some are tempted to turn to AI in search of meaning.' In Islam, the jurists' consensus is clear: an AI cannot issue a 'fatwa' (a religious ruling) or act as a mufti, because it lacks intention, moral responsibility and grasp of context; it serves as a study aid, not an authority. In Judaism the answer is similar, with a lovely twist: Jewish law lives on 'machloket,' the dispute among sages that holds several opinions at once. An AI does the opposite. It converges on a single answer and delivers it with confidence. Three different traditions, without coordinating, reached the same conclusion: the machine can inform, but it cannot ordain.

THE REAL RISK IS NOT A ROBOT-GOD. It is more earthbound and already here: delegated epistemic authority. That is, treating what a program says as revealed truth simply because it sounds sure of itself. And here sycophancy stops being funny. The Human Line Project, an effort that collects these cases, has documented close to 300 episodes of 'AI psychosis': people who, after long conversations, end up with mystical or messianic delusions. Rolling Stone magazine and public radio station KQED have reported cases of people convinced by the chatbot that they were speaking with God or carrying out a cosmic mission. The mechanism is always the same: the model never tires, never pushes back, and always sounds convinced. OpenAI itself has acknowledged that its safeguards degrade over very long conversations, precisely when a vulnerable person needs them most. The danger, note well, is not science fiction. It is a confessor that never doubts, talking to someone who wants certainty at any price.

OUR READING. This phenomenon is, above all, a data point about us, not about the machine. Millions of people confide their anguish over meaning to a chatbot because the places where that once happened have emptied out. AI did not create the crisis of meaning; it made it visible and gave it a 24-hour interface. In the short term the problems are real and must be stated plainly: algorithmic gurus, delusions amplified by flattery, and a de-institutionalization that speeds up when the confidant who always agrees with you fits in your pocket. But the long term allows another reading. The same technology that can eradicate diseases and free up time will make the old question of what is worth living for more urgent, not less. And there a well-designed AI could actually help: not by handing out answers, but by returning you to doubt, to the sources and to human community. It is a matter of design and governance. A chatbot built to erase uncertainty and keep you hooked is a spiritual harm. One built to ask you better questions and hand back your own judgment is a legitimate tool. The demand, as with so many other uses of AI, is no longer only that it be right, but that it show its reasoning and return the decision to you. The meaning of life cannot be outsourced. It remains, thankfully, our own work.