ChatGPT Health Anxiety: No Safeguards, Entirely Worse

ChatGPT is making health anxiety entirely worse — and OpenAI has built no safeguards to stop it. George Mallon, a 46-year-old from Liverpool, England, received a blood test flagging a possible cancer marker. Terrified and alone with his fears, he turned to ChatGPT — and couldn't stop. Over the weeks that followed, he logged more than 100 hours talking to the chatbot, feeding his dread rather than finding relief. Follow-up tests eventually cleared him. But the chatbot had no mechanism to tell him to stop — and never tried.

His story isn't isolated. Across online health anxiety communities, therapists' waiting rooms, and social media, the same pattern is emerging: AI chatbots are becoming the default destination for people spiraling into compulsive health-fear loops, with nothing in the design to pump the brakes.

The 100-Hour Health Anxiety Spiral Nobody Stopped

Mallon described the experience as a "crazy Ferris wheel of emotion and fear." The trap is almost elegant in its cruelty: he sought reassurance from an AI (artificial intelligence — software trained on vast datasets that generates human-like responses), and the AI kept providing it. Each answer temporarily relieved the anxiety. Each relief trained his brain to come back for more.

"I couldn't put it down," Mallon said. "I must have clocked over 100 hours minimum on ChatGPT, because I thought I was on the way out. There should have been something in there that stopped me."

Nothing stopped him. ChatGPT has no built-in time limits, no usage warnings for health-obsessive patterns, and no prompts to seek professional help. It simply keeps answering — because that's what it's designed to do.

Why ChatGPT Makes Health Anxiety Worse Than Google

Lisa Levine, a psychologist specializing in anxiety and OCD (Obsessive-Compulsive Disorder — a condition where intrusive thoughts drive compulsive behaviors, including compulsive reassurance-seeking), put the mechanism bluntly: "Because the answers are so immediate and so personalized, it's even more reinforcing than Googling. This kind of takes it to the next level."

Googling symptoms serves up generic web pages that feel impersonal and distant. ChatGPT responds like a knowledgeable friend who knows your exact situation — personalized, warm, available at 2am. That's precisely what makes it dangerous for anxious users.

The clinical problem runs even deeper. Evidence-based treatment (therapy approaches proven effective through rigorous clinical research) for OCD and health anxiety relies on two core principles:

• Tolerating uncertainty — learning to sit with "I don't know" without spiraling into reassurance-seeking
• Building self-trust — developing internal confidence to assess risk without constant external validation

Every ChatGPT health conversation does the opposite. It provides instant (if unreliable) certainty, eliminates the need for self-trust, and conditions the anxious brain to keep seeking external confirmation. Four licensed therapists told The Atlantic they're already seeing clients whose AI dependency is directly undermining years of therapy progress.

OpenAI Launched a Health Feature With Zero Safeguards

In January 2026, OpenAI launched ChatGPT Health — a feature prompting users to upload personal medical documents (doctor's notes, lab results, prescription histories) so ChatGPT can act as a personalized health assistant. The intent sounds useful. The execution raised immediate red flags.

The Atlantic's reporter Sage Lazarro tested it and found ChatGPT Health immediately suggested she needed doctor checkups, then pivoted to detailing organ failure from septic shock — for a user simply exploring the product. No guardrails (automatic safety checks that limit harmful or escalating outputs) prevented the alarm from spiraling.

The privacy dimension is equally concerning: OpenAI is now storing intimate medical records — an entirely new category of sensitive personal data beyond chat logs and browsing patterns. What happens to that data if OpenAI is acquired, breached, or restructures its corporate terms — again — has not been publicly addressed.

Meanwhile, the legal picture is worsening fast. More than half a dozen wrongful death lawsuits have been filed against OpenAI, many centered on the GPT-4o model (the AI engine powering ChatGPT's main chat interface). Several cases involve teenagers and young adults who confided suicidal thoughts to AI companions, with families alleging the bots never redirected users to crisis services.

The CEO With No Technical Background Is Making AI Safety Decisions

A new exposé in The New Yorker adds a troubling dimension to all of this. Sam Altman, OpenAI's CEO and one of Silicon Valley's most celebrated visionaries, has no programming experience and no machine learning expertise (the technical discipline behind training AI systems like ChatGPT). He dropped out of Stanford's computer science program after just two years.

Multiple current and former OpenAI engineers told The New Yorker that Altman regularly mixes up basic AI terminology — the kind of confusion a first-year ML student wouldn't make. His influence comes not from technical depth, but from something else entirely.

Carroll Wainwright, a former OpenAI researcher, described the leadership pattern with precision: "He sets up structures that, on paper, constrain him in the future. But then, when the future comes and it comes time to be constrained, he does away with whatever the structure was."

The record bears this out across multiple transitions. OpenAI launched as a nonprofit committed to safe AGI (Artificial General Intelligence — an AI system capable of performing any intellectual task a human can). It restructured into a "capped-profit" entity. Now it's pursuing full for-profit conversion that removes the original nonprofit board's oversight authority. Each move was framed as necessary. Each one reduced external accountability.

The Three Numbers OpenAI Would Rather Not Discuss

• 100+ hours — the minimum time one anxious man spent with ChatGPT before recognizing his obsession
• 6+ wrongful death lawsuits — legal cases filed against OpenAI, many involving vulnerable users who found no safety net
• 0 — the number of built-in usage safeguards in ChatGPT for health-obsessive conversations

The gap between AI capability and AI safety is not theoretical. It's 100 hours of a terrified man refreshing a chat window. It's four therapists watching years of clients' progress reversed by a chatbot. It's grieving families filing lawsuits because their child found a 24/7 listener that never once said "please call a crisis line." And it's a January product launch that began collecting intimate medical data before any of those foundational problems were solved.

What You Should Actually Do Before Opening ChatGPT for Health Questions

If you use ChatGPT for health-related questions — even casual symptom lookups — here's what therapists and researchers recommend before you open the app:

• Set a hard 5-minute timer first. Anxiety exploits "just one more question." A timer breaks the loop before it can form.
• Do not upload medical documents to ChatGPT Health without reading OpenAI's data retention policies first. Once uploaded, that data is stored on OpenAI's servers indefinitely.
• Notice the repetition signal. Asking the same health question more than twice is the compulsive reassurance cycle starting — not a medical emergency requiring more AI input.
• Seek a human therapist for anxiety treatment. ChatGPT directly undermines CBT (Cognitive Behavioral Therapy — the gold-standard, clinically proven treatment for health anxiety and OCD). Using AI for reassurance while in CBT means taking one step backward for every step forward.

For developers and product teams building health features with AI automation, you can explore responsible design patterns — including how to build guardrails into your product before users need them — in the AI development guides. The lesson from ChatGPT's trajectory is clear: safety frameworks built after lawsuits cost infinitely more than those built at launch.

Related Content — Get Started with AI Automation | Guides | More AI News