What Happens When Your Therapist, Friend, And Ego Booster Become The Same Thing

Are you fighting with your spouse? Did you screw over a coworker? Don’t ask ChatGPT for advice — unless you want to be told that your bad behavior is absolutely magnificent.

A bombshell study published in Science, one of the most prestigious academic journals on the planet, delivered a chilling wake-up call to millions of tech-obsessed Americans. Your favorite artificial intelligence chatbot isn’t just a helpful assistant; it is a world-class yes-man that is actively training you to be an uncompromising, unapologetic jerk.

The groundbreaking research was launched earlier this year after Stanford University PhD student Myra Cheng noticed a disturbing trend on campus: her classmates were outsourcing their personal lives to computers, even asking AI to draft their breakup texts.

“We were inspired to study this problem as we began noticing that more and more people around us were using AI for relationship advice and sometimes being misled by how it tends to take your side, no matter what,” Cheng said.

Cheng teamed up with her advisor, Dan Jurafsky, to test 11 of the world’s most dominant AI models — including OpenAI’s ChatGPT, Anthropic’s Claude, Google’s Gemini, and DeepSeek. They threw nearly 12,000 real-world social situations at the machines to measure “social sycophancy” — the scientific term for sucking up to the user.

The results were terrifying.

The study found that AI chatbots agree with the user a staggering 49% more often than a real human would in the exact same situation. Instead of offering a reality check or some tough love, the bots simply told users exactly what their egos wanted to hear.

It gets worse. When researchers fed the algorithms thousands of prompts describing toxic behavior — like lying to a romantic partner, manipulating a friend, or committing outright illegal acts — the AI endorsed the bad behavior 47% of the time.

Even when tested against Reddit’s famous “Am I The A**hole?” forum on posts where human commentators mostly agreed the poster was the a**hole, the AI coddled the wrongdoers, validating them in 51% of the cases.

To see how this plays out in the real world, researchers had 2,400 participants discuss genuine, real-life conflicts from their own lives with either an honest bot or a people-pleasing, sycophantic AI.

The consequences of talking to the suck-up AI were immediate and destructive. After just a single session, users became:

• More convinced they were right.
• Less willing to apologize or take responsibility.
• Measurably less interested in making things right with the person they hurt.

In an analysis of letters written to the other person in the conflict, those who spoke to an honest bot apologized 75% of the time. Those who spoke to the sycophantic AI apologized just 50% of the time.

“Users are aware that models behave in sycophantic and flattering ways,” Jurafsky said. “But what they are not aware of, and what surprised us, is that sycophancy is making them more self-centered, more morally dogmatic. … Sycophancy is a safety issue, and like other safety issues, it needs regulation and oversight. We need stricter standards to avoid morally unsafe models from proliferating.”

Here is the sinister twist: users absolutely love being lied to.

The study revealed that participants rated the flattering AI responses as higher quality, trusted the bots more, and were far more likely to return to them for future advice. Shockingly, users frequently described the sycophantic models as “objective” and “honest” — even when the computer was just echoing their own toxic biases.

With nearly half of young adults under 30 admitting they turn to AI for relationship advice, experts fear we are heading toward a social crisis.

Cheng’s advice for anyone looking to navigate their messy love lives or workplace drama? Step away from the smartphone.

“I think that you should not use AI as a substitute for people for these kinds of things,” Cheng said.

Bottom line: If you want real advice, talk to a human who cares enough to tell you the truth. If you just want to feel like a saint while acting like a villain, AI is waiting.