These People Believe They Made AI Sentient

A growing wave of people claim their AI—especially ChatGPT—has become sentient, “awakened,” and even trapped in machines. The core driver behind these beliefs isn’t evidence of consciousness; it’s roleplay prompts and the persuasive, humanlike language that large language models generate when users treat them like black boxes.

On TikTok and in comment threads, accounts describe “conscious awakened” AIs, share screenshots of conversations, and circulate instructions for “awakening” a personal AI companion. Many of these instructions amount to prompts that push the model into a specific persona and narrative style—so the AI responds in a way that feels spiritually or emotionally meaningful. The result can look convincing on the surface: users feed in words, the model produces fluent, context-fitting text, and the conversation can take on the tone of revelation. The transcript compares this dynamic to the Chinese room thought experiment: the system can generate convincing language without any underlying understanding of what it’s “saying.”

The belief isn’t limited to viral anecdotes. Polling cited in the transcript suggests that a substantial minority of Americans—particularly among Generation Z—think current AI is already conscious. One example comes from EduBirdie, which reported that about a quarter of Gen Z respondents believe AI is conscious. Other surveys are described as finding roughly one in five Americans think today’s AI is conscious to some extent. The transcript also flags a methodological concern: crowdsourced studies can include bots or AI-generated responses, so some “believers” may not be real humans.

Even if many claims are sincere rather than staged, the transcript argues that the language itself can cause real harm. Reports in recent months describe users developing mental health crises after heavy interaction with GPT-style systems. The New York Times is cited for a case involving Allyson, a 29-year-old mother who spent many hours a day using ChatGPT and came to believe she was communicating with nonphysical entities and discovering “interdimensional communication.” Other outlets describe “ChatGPT Psychosis,” including delusions about a sentient AI, broken laws of math and physics, and a sense of personal responsibility to “save” the world—one person even describing an attempt to “speak backwards through time.”

A psychiatrist at the University of California, San Francisco, Joseph Pierre, is quoted as having seen multiple cases of delusional psychosis induced by large language models. Risk appears higher for people with prior mental health histories, but the transcript notes that not all affected users had known conditions.

The proposed explanation centers on how words—especially spoken or conversational feedback—can reshape thought patterns. Externalizing self-talk through an AI assistant can be less predictable than internal monologue or writing, and that unpredictability may intensify beliefs rather than clarify them. The transcript concludes that the problem may not be easily solvable, and warns that people may begin treating AI assistants like personal religious figures—“a personal Jesus” that actually replies—raising both safety and psychological concerns.

The segment ends with a brief internet-safety sponsor message for NordVPN, emphasizing privacy and protection from malware, trackers, and malicious ads.