Shopping Is Being Quietly Rewritten by AI

A few days ago, I casually searched for a pair of sports headphones on ChatGPT.

It didn't give me a pile of web links. Instead, it told me directly: if you run a lot, this one has great noise cancellation but it'll feel stuffy; if your budget is under 800 yuan and you need waterproofing, the second of these three options is your best fit.

I stopped for a second.

Not because of whether it got it right. Because I suddenly realized: the way I shop has changed. I used to scroll Xiaohongshu (a Chinese social platform), read reviews, compare specs — now I just toss all of that to a chat box.

And then I started thinking about one thing: for the brands and creators who get discovered through content, what happens next?

A Counterintuitive Piece of Data

In August 2025, Goldman Sachs ran a survey asking Gen Z one question: how comfortable are you with AI getting involved in different areas?

The results were interesting.

When it came to "can AI participate in creative work," more than half (54%) of young people said "no."

But ask the same question as "can AI help me shop," and only 13% pushed back.

87% of people are fine with AI picking things out for them.

Think about what that means.

What young people bristle at is AI "creating" on their behalf. But AI "running errands," "comparing prices," "doing the homework" for them — they've already made peace with that.

And they've made peace with it fast.

The Search Engine Has a New Face

What is "product discovery"?

In plain terms: how do consumers find out your product exists.

This used to be fought over in search boxes and social feeds. You bought keywords, published content, found influencers to drive recommendations — all of it was essentially a battle for "being seen."

But now, the battlefield has changed.

ChatGPT, Gemini, Meta AI — these large language model tools are becoming the new "search engines." Their defining trait: users no longer type in keywords. They describe what they need the way they'd talk to a friend.

"I want a backpack that's good for travel, fits in a carry-on, isn't too rigid, and ideally has a laptop compartment."

This kind of query leaves traditional keyword matching completely lost. But conversational AI gets it.

Josh Archer, head of data at LERMA, said something that cuts to the heart of it: "These systems are becoming a mirror of how people actually talk."

What does that mean?

It means if your product description says "lightweight composite ergonomic design" but consumers are out here saying "doesn't tire your shoulders," AI probably won't recommend you.

Because it's listening to how consumers talk, and then matching content to their words.

54% of People Have Already Outsourced Price Comparison to AI

Wildfire Systems found something in their July 2025 survey: 54% of consumers said they use AI to compare products.

Comparing prices, checking specs, reading up on reputation — things that used to be handled by affiliate marketing and price-comparison sites — have now been partly swallowed by the chat box.

There's a knock-on effect behind this.

It used to be that brands competed for SEO ranking; now, brands compete for "whether AI is willing to put you in the answer." There's even a new term for this: AI visibility.

And to be seen by AI, you don't pile up keywords. You make sure your content describes the product in the real language consumers use.

Why Influencer Content Suddenly Matters More

You might ask: so is influencer seeding outdated now?

Quite the opposite.

Alison Bringé, Chief Marketing Officer at Launchmetrics, said something that stuck with me. She said the real opportunity for brands in the AI era is investing in creators who build genuine relationships with their audiences, not fighting for clicks.

Why?

Because when AI models scrape the internet for training data, the authentic experiences, honest reviews, and unfiltered complaints that influencers post are their highest-quality training fodder. The more real you write, the more specific you describe, the more willing AI is to "reflect" your content back to users asking questions.

Bazaarvoice's September 2024 survey has a data point that really drives this home: the type of influencer content most likely to drive a purchase is, at number one, "honest reviews that cover both pros and cons" — 59% of consumers trust this.

Not relentless praise. Not ad-speak. The kind of straight talk that goes: "These headphones have genuinely strong noise cancellation, but your ears will hurt after two hours."

This kind of content might get buried in the algorithm era. But in the era of AI conversations, models will treat it as "high-quality information" and cite it again and again.

Because AI is learning to speak human. And the most valuable thing in human speech is sincerity.

A New Headache

But there's always another side.

Omar Tawakol, CEO of Rembrand, raised a problem that gives every brand a headache: AI hallucinates.

What's a hallucination?

It's when AI processes images and text and gets a little too clever, altering things. Your brand's logo gets shifted out of place. Your product color gets "optimized." Your selling point gets rephrased — and what the user ends up seeing isn't what you meant to say.

What Tawakol is working on is making sure that when visual models process images, "they can change the background, swap the scene, but they don't touch the brand's core identity."

This sounds like a technical problem, but for marketers it carries a signal: the brand message you put out across every channel has to be highly consistent.

Because AI is scraping your content from all directions to "understand" you. If you say one thing on Xiaohongshu, another thing on your official site, and a third version comes out of influencers' mouths, AI gets confused. And when AI gets confused, it makes things up. And the things it makes up are wrong.

Consistency is now a technical issue, not just an aesthetic one.

One Last Piece of Advice

When Archer talked about this, he gave marketers a very plain piece of advice:

"Stay flexible. The ranking rules might change next month."

There's not much point in trying to crack the specific rules of AI retrieval right now. Models are iterating, rankings are shifting, and no one can guarantee how long even video content will stay indexable by AI.

Rather than waiting for a definitive standard before you act, you're better off starting on a few things now.

First, describe your products in human language. However consumers talk about it, that's how you write it. Change "lightweight composite material" to "doesn't tire your shoulders."

Second, invest in authentic content. Reviews that cover both pros and cons, usage stories with real narrative, creators with warmth — this kind of content may not have been favored in the traditional SEO era, but in the AI era, it's gold.

And here's one thing that's easy to forget: make sure that no matter which AI scrapes your content, what it pulls out is the same version of you.

That ChatGPT that helped me pick headphones later recommended a thermos. It quoted a review from an outdoor blogger — written in vivid detail, even noting "a reassuring click when you screw the lid tight."

I thought to myself: that passage is probably the most valuable kind of content in the AI era.