There is finally a clinical term for the thing you do when ChatGPT gives you an answer and you just go with it. Researchers are calling it "cognitive surrender." And a new study with over 1,300 participants proves it is not just laziness. It is a measurable, predictable collapse in human reasoning.

The numbers are brutal. Across more than 9,500 individual trials, subjects accepted faulty AI reasoning 73.2% of the time. They only overruled the AI 19.7% of the time. The AI was wrong in these tests. Deliberately, provably wrong. And nearly three out of four people went along with it anyway.

The researchers found that AI outputs written in fluent, confident language were treated as "epistemically authoritative." In plain English: if the chatbot sounds smart, people stop checking whether it is actually right. The confident tone lowers your mental defenses. It shuts off the part of your brain that would normally say "wait, let me think about this."

Not everyone fell for it equally. People who scored high on fluid IQ tests were significantly less likely to be misled. They questioned the AI more, overruled it more, and relied on it less. On the other end, people who started out trusting AI were far more likely to be led astray. High trust plus faulty AI equaled the worst outcomes.

Here is the part that should worry you. The researchers point out that cognitive surrender is not inherently irrational. If the AI is actually right most of the time, deferring to it is a reasonable strategy. The problem is that it creates a structural dependency. Your reasoning becomes exactly as good as the AI you are using. No better. If the AI is brilliant, you look brilliant. If the AI hallucinates, you hallucinate with it.

This connects directly to the BCG study from last week that found AI "brain fry" among workers who overuse AI tools. The productivity revolution has a dark side nobody planned for. The tools are not just doing the work. They are replacing the thinking. And once you outsource your reasoning to a machine, getting it back is not as simple as closing the tab.

The implications for everything from education to legal work to medical diagnosis are enormous. Courts are already sanctioning lawyers who submitted AI-hallucinated briefs. Doctors are using AI to screen patients. Students are using it to write papers. In every case, the question is the same: is the human still thinking, or have they surrendered?

As the researchers put it: "As reliance increases, performance tracks AI quality, rising when accurate and falling when faulty." That is a polite way of saying we are building a civilization-scale dependency on systems that are confidently wrong about 20% of the time. Let the prompter beware.