Every "Just What You Wanted" You Scroll Past — Behind Every One Is the Same Thing

A few days ago, I casually tapped on a pair of running shoes in a shopping app.

Didn't buy them.

Closed it.

The next day I opened the app again, and running shoes were all over the homepage. The day after, WeChat Channels (Tencent's short-video feed) started pushing running tutorials at me. The day after that, I got a text: "The running shoes you're watching have dropped in price."

I was a little uncomfortable — and a little impressed.

Uncomfortable, because I felt an invisible hand pushing me along. Impressed, because that hand had uncanny aim.

Have you ever wondered who that hand actually is?

It has a name: AI personalization.

What Is AI Personalization, Exactly?

Let me break the term down for you first.

Personalization, at its core, means "I know who you are, so I show you what you want to see."

So what's AI personalization? It means that on the "I know who you are" step, the machine took over for humans — and it does it faster, finer, and on a scale the human mind can't even begin to imagine.

What you've clicked, how many seconds you lingered, where you swiped away, when you placed an order, when you abandoned one — these fragments are impossible for a human brain to piece together, but AI can. And once it does, it paints a portrait of you that understands you better than you understand yourself.

In academia, this has an acronym: AIP (AI-enabled personalization).

This Hand Reaches Into Every Step of Your Buying Journey

A lot of people think personalization just means a few product recommendations on the homepage.

That's wildly underestimating it.

Marketing thinkers have broken down the entire process — "from seeing an ad, to buying, to coming back to repurchase" — into five stages, and they call it the customer journey. AI personalization is busy at every single stop along the way. Let me walk you through them one by one.

Stop one: You've never even heard of this brand.

What's AI doing right now? It's building a picture of you.

Based on your browsing, searches, location, and spending habits, it tags you. You might be slotted into "25–35 years old, health-conscious, price-sensitive."

There's a name for this: personalized profiling.

Sounds pretty harmless. But think about it — before the brand has even met you, it already has you figured out. Every ad you see from here on is tailor-made for "people like you."

Stop two: You're getting interested, and you tap in.

What AI does here is called navigation.

It decides which products to show you, in what order, and which one goes on top. You think it's "sorted by sales" — actually, it's "sorted by how likely you are to place an order."

Two people open the same app at the same time and see completely different homepages. The shelf you're browsing has been arranged just for you.

Stop three: You're a little tempted, but you haven't placed the order yet.

Now AI starts pushing you — or, in the academic term, nudging you.

A countdown on a limited-time discount. An "only 3 left" stock alert. That little line that reads "customers who bought this also bought." You think you decided to buy on your own — every step of the way, someone has been gently tapping you on the back.

Stop four: You've made the purchase.

You think it's over? It isn't.

AI starts calculating whether you'll come back. It watches your repurchase cycle, your satisfaction, your churn risk. If you haven't returned in three months, it will drop a coupon on you with surgical precision — right around the time you're about to forget this brand exists.

This is called retention personalization.

Stop five: You've become a loyal customer.

This is where AI does its most hidden work — it's deciding how to treat you.

Same VIP tier, and one person gets a birthday gift box while another gets a "double points" text. Why? Because AI has figured out that you're sensitive to "being remembered" — but numb to "saving money."

See — from before you ever walked through the door to the point where you're already an old regular, that hand has never once let go.

But Then Things Start to Go Sideways

Earlier I said I was "impressed." The more I sat with it, the more something felt off.

Push personalization to its extreme, and it runs into a wall.

Wall one: Is the goal to "get me," or to "trap me"?

The more accurate the profile, the more accurate the recommendations; the more accurate the recommendations, the narrower the world you get to see.

There's a term for this: the filter bubble.

You click on running shoes once, and you see running shoes forever. You click on a conspiracy theory once, and you see conspiracy theories forever. AI won't show you things you don't like — because its goal is to make you linger, click, and buy more, not to show you a complete picture of the world.

In the short run, that's the commercial optimum. In the long run, it's feeding a version of you that gets more and more fixated.

Wall two: You don't get a say in your own profile.

The portrait AI paints of you — you can't see it, and you can't change it.

It might tag you as a "low-value user," so you always get the worst offers. It might label you as "impulsive," so you get nothing but limited-time discounts. You don't know, and you can't appeal.

In academia, there's a term for this: profile opacity. In plain English: the machine has labeled you, and you don't even have the right to be told.

Wall three: When it gets too accurate, it gets creepy.

Has this ever happened to you — you mention something out loud to a friend, and the moment you open your phone, you see an ad for it.

A chill runs down your spine.

Even if there's a technical explanation (your earlier behavior already gave away your interests), that "being eavesdropped on" feeling is real unease. Personalization has a tipping point — cross it, and the user's experience flips from "thoughtful" to "surveilled."

Where exactly that tipping point sits, no one can give you a precise number. But what happens after you cross it is clear enough: users start turning off recommendations, clearing their browsing history, and uninstalling the app.

So What Do We Do?

I'm not going to hand you a bowl of "personalization is the future" chicken soup, and I'm not going to chant a "personalization violates privacy" slogan.

It's not that black and white.

The more clear-headed approaches I've seen basically fall into three directions.

First, give the choice back to the user.

Let users see their own profiles, edit them, and turn them off. This sounds like it would break personalization — in practice, the users who choose to stay end up trusting you more.

Second, draw a "do-not-cross" line for personalization.

Not every piece of data should be used. Not every moment is the right moment to push. When a user has just suffered a setback, just been laid off, just lost a loved one — that's when push notifications need to show restraint.

The real rival of personalization isn't "not accurate enough" — it's "too accurate."

Third, accept that personalization will fail.

No matter how precisely AI calculates, it can't account for a user who had a fight with his wife this morning and is in a lousy mood and doesn't want to buy anything. It can't account for a user who's actually buying on someone else's behalf. It can't account for a user who's just browsing and has no money to spend at all.

Personalization is probability, not destiny. Treat it like a god and it will disappoint you. Treat it like a tool and it will help you.

One Last Thing

Let's come back to that hand from the beginning.

It isn't evil. It makes shopping smoother, routes information to the right place, and helps a lot of genuinely good things — the kind that would otherwise be drowned out — find the people they were meant for.

But it isn't neutral, either. It's quietly reshaping the world you see, the decisions you make, and the things you think you want.

What worries me more is this — most people have no idea this hand even exists.

That thing you bought today — did you really want it? Or was it the thing some model calculated you'd be "highly likely to order"?

That question is worth sitting with.

Even if thinking about it doesn't yield an answer. At the very least, the next time you see a "just what you wanted" recommendation, you'll pause for an extra half-second.

That half-second is the distance between "being personalized-to" and "seeing personalization at work."