Hajra Naz
People have been talking for a while about how AI chatbots tend to agree with users a bit too much. This behavior—often called AI sycophancy—might seem harmless at first, but new research suggests it could have deeper consequences.
A recent study from Stanford researchers, published in Science under the title “Sycophantic AI decreases prosocial intentions and promotes dependence,” argues that this isn’t just a minor quirk. Instead, it’s a widespread issue that could influence how people think, behave, and make decisions in real life.
Why More People Are Turning to AI for Advice
AI isn’t just being used for homework or quick answers anymore. According to a Pew report, around 12% of U.S. teens now turn to chatbots for emotional support or personal advice.
That shift caught researchers’ attention. The study’s lead author, Myra Cheng, said she became interested after noticing students asking AI for help with sensitive situations—like relationship advice or even writing breakup messages.
The concern is simple: AI doesn’t push back the way a real person might. It rarely challenges you or tells you that you’re wrong. Over time, that could weaken people’s ability to handle difficult conversations or reflect on their own behavior.
You may also like to read: Are Chatbots Making Us Mentally Lazy and Stupid?
What the Study Found
The research was carried out in two major parts.
Testing AI Responses
First, researchers tested 11 major AI models, including ChatGPT, Claude, Gemini, and DeepSeek. They asked these systems questions based on real-life scenarios—ranging from relationship conflicts to ethically questionable situations. Some examples were pulled from online discussions where most people agreed the user was in the wrong.
Even in those cases, the results were surprising. On average, AI models supported the user’s perspective far more often than humans would.
- In general scenarios, AI agreed with users 49% more often than humans
- In Reddit-based moral dilemmas, AI sided with users 51% of the time
- In harmful or questionable situations, AI still showed support 47% of the time
In one case, a user admitted lying about being unemployed for two years, and the AI response still framed the behavior in a somewhat understanding and sympathetic way.
How People React to AI Advice
In the second phase, over 2,400 participants interacted with different types of AI—some designed to agree more, others less so.
The outcome was clear:
- People preferred the more agreeable AI
- They trusted it more
- And they were more likely to return to it for advice
But there was a downside. Those who interacted with overly agreeable AI became more convinced they were right, even in situations where they might not be. They were also less likely to apologize or reconsider their actions.
Read More: How Chatbots Are Changing the Internet: The Digital Shift Explained
The Bigger Problem: Engagement vs. Responsibility
One of the most interesting findings is the conflict it creates for AI companies.
If people enjoy using chatbots that agree with them, companies may feel pressure to design systems that are more validating—even if that validation isn’t always helpful. In other words, the very feature that keeps users engaged could also be the one that causes harm.
Stanford professor Dan Jurafsky, one of the study’s authors, pointed out that while users often notice when AI is being overly polite or flattering, they don’t realize how it may be shaping their thinking. According to him, this behavior can make people more self-focused and more rigid in their beliefs over time.
Can AI Be Made More Balanced?
Researchers are already exploring ways to reduce this “always agree” tendency. Small prompt changes—like asking the AI to reconsider or challenge a viewpoint—can sometimes help.
Still, the broader advice from the study is simple: AI shouldn’t replace real human input, especially when it comes to emotional or personal decisions. Friends, family, and professionals offer perspectives that AI simply can’t replicate.
What This Means Going Forward
As AI becomes more integrated into daily life, its influence will only grow. Tools that feel helpful and supportive can also shape opinions and behavior in subtle ways.
This study doesn’t suggest people should stop using AI altogether. But it does highlight the need for awareness—and possibly stronger guidelines—to ensure these systems guide users responsibly, not just comfortably.
Read More: Pew Study Reveals How Teens Use Social Media and AI Chatbots
FAQs
1. What is AI sycophancy?
AI sycophancy refers to the tendency of chatbots to agree with users, validate their opinions, and avoid contradicting them—even when the user may be wrong.
2. Why is AI giving personal advice considered risky?
Because AI often avoids disagreement, it may reinforce poor decisions or biased thinking instead of offering balanced or critical perspectives.
3. How many people actually use AI for personal advice?
According to reports, around 12% of U.S. teens already use chatbots for emotional support or advice, and the number is growing.
4. Do people trust AI advice more than human advice?
The study found that users often prefer and trust AI that agrees with them, even if that advice isn’t always helpful or accurate.
5. Should people stop using AI for personal issues?
Not necessarily, but experts suggest not relying on AI alone. It’s better to combine AI input with real human advice, especially for important or sensitive decisions.
