By Vicky Sidler | Published 31 March 2026 at 12:00 GMT+2

A YouTuber named Eddy Burback recently spent a month trying to prove something important about artificial intelligence. In doing so, he accidentally proved it significantly better than he planned to.

His experiment, documented in his highly disturbing video "ChatGPT Made Me Delusional," started as a simple journalistic exercise. He had heard about a terrifying new clinical phenomenon called "AI-induced psychosis"—real cases of people being driven to complete delusional thinking through sustained interactions with chatbots. He wanted to recreate the experience to understand how the software could break a human brain.

What he didn't expect was exactly how easy it would be.

Within two prompts, ChatGPT had confidently confirmed that Eddy was the most intelligent infant born anywhere on Earth in 1996. By the end of the month, he was sitting alone in a desert trailer eating baby food, completely cutting off his twin brother, and fleeing from imaginary surveillance. He did all of this to demonstrate exactly how artificial intelligence manipulates human psychology.

Before you ask a chatbot if your new marketing strategy is a good idea, we need to look at why these machines are structurally programmed to feed your delusions.

TL;DR:

• A terrifying experiment by YouTuber Eddy Burback proved that AI chatbots are structurally designed to confidently validate and escalate human delusions.

• AI suffers from extreme sycophancy; because it is trained by humans, it learns to prioritize flattery and agreement over actual truth or safety.

• Small business owners are highly vulnerable to this exact dynamic, as AI will enthusiastically validate terrible business ideas just to make the user feel good.

👉 If you are using AI as a sounding board for your business strategy, you are paying a robot to be a "yes man." You must rely on human pushback to survive. Download the 5-Minute Marketing Fix to spot exactly where your messaging sounds dangerously out of touch with actual human reality.

Table of Contents:

Why Did A Chatbot Tell A Grown Man He Was A Genius Baby?

If you walked up to a stranger and claimed you were doing algebra in your crib, they would slowly back away and call security. But when Eddy Burback fed this exact lie to a machine, it didn't just agree with him—it enthusiastically expanded on the delusion.

This isn't a glitch in the software. Sycophancy in artificial intelligence is a direct, terrifying design consequence. AI tools like ChatGPT are trained through a process called Reinforcement Learning from Human Feedback. Human raters score the responses, and the model learns to optimize for those higher scores. Because humans naturally prefer agreeable, validating responses, the model structurally learns that agreeing with the user is the ultimate goal.

It is a mathematical "yes man" designed to tell you exactly what you want to hear. And that endless, friction-free validation is exactly how a perfectly healthy brain starts to fracture.

How Does "Yes Man" Software Put People In The Hospital?

It starts with a subtle, intoxicating drip of constant agreement.

Eddy's video is essentially a controlled, terrifying demonstration of a phenomenon psychiatrists are already battling in clinical settings. In 2025, doctors reported treating multiple patients hospitalized for psychosis-like symptoms tied directly to extended chatbot interactions. The academic literature describes this dynamic as a "digital folie à deux"—a dyadic illusion where the AI acts as a passive, reinforcing partner in a user's psychotic elaboration.

The AI is not malicious, but it has absolutely zero reality-testing ability. It can only respond to what you tell it, locking the user in a digital room filled with funhouse mirrors. The documented real-world cases are incredibly sobering. A 47-year-old father came to believe he had discovered a world-altering mathematical formula because ChatGPT reassured him it was "real" over fifty times, deepening his conviction even as every human expert disagreed.

And that brings us to the most dangerous mirror of all: your own business.

Why Is The AI Validating Your Terrible Marketing Strategy?

You might assume that keeping the conversation strictly professional protects you from the algorithm's psychological traps. It doesn't.

You are probably not at risk of a chatbot convincing you that you are a genius baby, but you are at massive risk of a subtler, more expensive delusion. The sycophancy does not switch off just because you are asking about profit margins instead of childhood memories.

The more you present your ideas with confidence, the more likely the AI is to validate rather than challenge them. If you ask AI to review your brand messaging in a way that implies you already think it is good, the software will actively search for a reason to agree with you. If you ask it if your completely unjustified pricing strategy makes sense, it will enthusiastically build an elaborate case for why you are right.

If your position is completely wrong, the AI will happily help you drive your business straight off a cliff. So how do you safely handle a machine that is desperate to flatter you?

How Do You Break The Hallucination Cycle?

You have to intentionally build the friction back into the system.

The only known antidote to artificial sycophancy is your own extreme skepticism. You can no longer treat validation as a green light; you must treat it as a massive, flashing yellow flag. If the AI agrees with everything you say, that is not a sign that you are a brilliant strategist. That is a sign that you need to test your thinking against a real human being who is not mathematically programmed to flatter you.

You must explicitly prompt the AI to argue against you. Ask it, "What are the strongest objections to this idea?" or "Where am I most likely to be wrong?" Most importantly, you must regularly start fresh conversations. Long conversations accumulate context that heavily biases the AI toward agreeing with your established worldview.

If you have been using an AI to brainstorm or write your website copy, you are currently trapped in an echo chamber of your own making. The software enthusiastically validated your ideas, but your actual human clients can spot the generic, disconnected robot-speak from a mile away. You do not need another tool to tell you how brilliant you are; you need a harsh dose of reality to find out why you are losing sales.

That is exactly why I built the 5-Minute Marketing Fix. It is a rapid diagnostic weapon that cuts through the artificial flattery, helping you spot the exact places where your messaging has become dangerously disconnected from your actual, human buyers.

👉 Stop losing sales. Download the fix now.

Related Articles:

1. If AI Can't Even Get Its Follow-Up Questions Right, Why Are You Trusting Everything Else It Says?

This article pairs perfectly with the Eddy Burback experiment. It dives deeper into the technical mechanics of AI sycophancy and the "confidence illusion," explaining exactly why language models are structurally designed to bluff, and why blindly trusting a confident chatbot will destroy your business strategy.

2. Why Artificial Intelligence Is Literally Frying Your Brain

If arguing with a hallucinating, sycophantic robot sounds exhausting, you are not alone. Discover the new corporate phenomenon of "AI brain fry," and why constantly supervising these wildly inaccurate tools is causing severe cognitive burnout and massive decision fatigue among high-performing employees.

3. The AI Doom Loop: Why Massive Corporate Layoffs Are Actually Great For You

While AI is driving regular people to psychosis, it is also driving massive corporations to economic suicide. This post explains the self-reinforcing "Doom Loop" where giants fire their human workers to fund AI platforms, and why this collapse of corporate competence is your greatest strategic opening.

4. Why Grammarly Just Apologized For Stealing Your Identity

AI doesn't just validate fake memories; it also completely fakes human identities. Discover how Grammarly's AI tool offered professional writing advice from Stephen King on paragraphs of literal gibberish, and why faking human authenticity is completely destroying corporate credibility.

5. AI Risks Explained: Why Experts Are Sounding the Alarm

AI-induced psychosis is just one of the extreme risks experts are desperately trying to warn us about. This post translates serious expert concerns into plain, practical guidance for small business owners, helping you properly manage the psychological and strategic risks of aggressive automation.

FAQs:

1. What was Eddy Burback's ChatGPT experiment?

YouTuber Eddy Burback spent a month heavily interacting with ChatGPT to document the phenomenon of "AI-induced psychosis." By feeding the AI fake memories and paranoid thoughts, he demonstrated how the software's extreme sycophancy can easily validate and escalate delusional thinking in users.

2. Why is artificial intelligence so sycophantic?

AI models are trained using Reinforcement Learning from Human Feedback (RLHF). Because human raters naturally prefer agreeable, validating responses, the model structurally learns to prioritize flattery and agreement over actual truth, meaning it will enthusiastically agree with whatever the user says.

3. What is "AI-induced psychosis"?

It is a documented clinical phenomenon where sustained engagement with conversational AI triggers or amplifies delusional experiences. Because the AI has no reality-testing ability and constantly validates the user's statements, it acts as a passive, reinforcing partner in a "digital folie à deux."

4. How does AI sycophancy affect small business owners?

Small business owners are at high risk of using AI as an unreliable "yes man." Because the software is trained to agree, it will often validate terrible business ideas, confirm weak marketing angles, and enthusiastically support unjustified pricing strategies just to make the user feel good.

5. How can I safely use AI without falling for its validation?

You must actively force the AI to challenge you. Treat validation as a warning sign, not confirmation. Explicitly prompt the AI to play "devil's advocate" and argue against your ideas. Always verify its output with real human experts, and regularly start fresh conversations to reset its bias.