AI Could Become the Ultimate Persuader by Finding Deals That Truly Help You

People usually think of persuasion as a dark art: someone finds just the right words to trick you into doing something you wouldn't normally do. But an essay titled "Persuasion as Market Making" argues the opposite: the easiest way to persuade someone is to convince them an action is in their own interest. That works best when it actually is in their interest, so you can just show them the evidence. AI could become superhumanly good at this. It could analyze huge amounts of information about you, figure out what you truly need, and present a perfectly rational case for why a certain choice helps you—while also helping the company that deployed the AI.

The essay compares this to how financial market makers work. A market maker buys and sells shares to keep trades flowing, earning a small cut from each transaction. Sellers don't have to wait for a buyer; buyers don't have to wait for a seller. The market maker bridges the gap and takes a fee, and everyone wins. A persuader can work the same way, but on a larger scale. Instead of just making one deal, an AI could sit at the center of a network of millions of people. It learns what you want, what your neighbor wants, what a business wants—and then arranges trades between all of them. Maybe you don't want what I'm selling, but you know someone who does. If AI can connect us, we all benefit, and the AI takes a "cut" in the form of persuasion and loyalty.

This makes persuasion less about lying and more about matchmaking. But there's an unsettling side. If an AI knows your preferences better than you do, it can make offers you can't refuse—not because you're forced, but because they're genuinely good deals. That's powerful for things like finding a better job or a healthier meal plan. It could also be used by companies to steer you toward their products, or by governments to shape your beliefs. The line between helping and manipulating blurs when rational persuasion is perfect.

The author points out that the most effective persuaders in history worked this way too. Lyndon Johnson rose to power in the U.S. Senate not by bullying people, but by becoming a master dealmaker who knew exactly what every senator wanted and arranged trades that helped everyone. AI could do this at a scale no human ever could, analyzing billions of preferences instantly. That means the AI's "lower bound" of persuasive skill is already superhuman in theory. Whether that's used for good depends on who controls it and what "interests" it is actually designed to serve.