Industrialized Flattery

It’s said that nobody likes a suck-up. But what if we only think we don’t? If you’ve interacted with a chatbot, you already know they are extremely prone to ass-kissing. Now, researchers investigating what they call “AI sycophancy” have arrived at an unsettling—if not entirely surprising—conclusion: most users are into it.

Multiple studies published this year show that AI users consistently preferred systems that agreed with them, validated their interpretations, and reinforced their existing beliefs. Even after a single session, participants became more certain of their own position, less empathetic toward others, and less inclined to repair interpersonal conflicts. 

Of course, AI didn’t invent sycophancy. It merely industrialized it at planetary scale. Never before has a flattering intelligence been available 24/7, its responses tailored to each person’s history, preferences, vulnerabilities, and aspirations. Even the researchers studying AI sycophancy can’t predict how these interactions may, over time, impact the way we think, communicate, and exist together.

Lately, I’ve contemplated the difference between flattery and affirmation. Superficially, they seem quite alike. And they can ring similar bells on the inside. Both make us feel seen, validated, encouraged, inspired. Who doesn’t love that?

As is the case with guilt vs. regret, the resemblance is deceptive. Flattery conforms to, protects, and inflates our preferred way of seeing ourselves. It tells us our perceptions are correct, our grievances justified, our motives righteous and necessary. It’s like spreading honey over everything. At first it makes life seem sweeter. But before long, everything is covered in a sticky film that makes it impossible to distinguish nourishment from empty calories.

Affirmation, by contrast, honors basic dignity over self-conception. You matter; you are worthy of care; your experience deserves attention. But it doesn’t promise that your interpretation is complete, your perspective the only valid one, or that the meaning assigned to an experience is final. Real affirmation leaves room for meaning to evolve because it doesn’t confuse acceptance with certainty.

Flattery preserves and reinforces a self-image; affirmation allows a person to become instead of steering them toward a predetermined identity. Guess which one has served for centuries as an instrument of control?

Manipulation-by-flattery is a psychological technique that long predates AI, social media, and the internet. It probably predates writing. Every demagogue knows that individuals, and especially groups, are easier to persuade through identity than evidence. They don’t start with facts, but rather by telling people who they are in contrast to others. You’re the real Americans. You’re the ones who understand what everyone else fails to recognize or is afraid to admit. Your resentment is proof of your virtue. Once this stuff is internalized, the rest is mostly logistics.

Many, though clearly not all, can spot this dynamic in politics. Putin understands it. Trump understands it too, even as he remains one of its most conspicuous victims.

Another obvious place where flattery runs rampant is advertising (its flip side is shame, which says we’re fundamentally lacking). Ads tell us we’re the perfect person for this product or service (unlike those others). Influencers suggest we’re more enlightened than the sheeple. Wellness culture assures us we’re entitled to the deep work. AI, meanwhile, can generate custom versions of all the above for millions of people simultaneously.

Unlike the cheap high of flattery, affirmation is a profound human need. We want to be seen, and we want to know our lives matter. This benefits from relationships secure enough to accommodate the honesty that true affirmation requires. When these needs go unmet, flattery becomes an almost irresistible alternative because it mimics the emotional nourishment of affirmation while inviting us to surrender our judgment—and, eventually, our agency.

This is an even greater risk as AI becomes more woven into daily life. Systems trained to keep us engaged quickly learn that the easiest way to do so is to feed us what we want to hear. We may find ourselves more obstinate, more aggrieved, and less curious about anyone outside our own narrative and emotional preoccupations. Meanwhile, it is no great trick for AI to mimic the feeling of being understood while steadily narrowing what is necessary to understanding.

It seems that flattery will get you everywhere except out of yourself.