Meta's AI Shopping Assistant Is Solving the Wrong Problem

On March 3, Meta rolled out an AI shopping assistant across its apps in the US. You can ask it for product recommendations in a chat interface, and it'll surface options from brands on Facebook and Instagram.

The coverage was what you'd expect. New feature, here's how it works, here's where to find it. Nobody asked the one question that actually matters: where do those recommendations come from?

Follow the Money

Meta made $164.5 billion in revenue in 2025. Over 97% of that came from advertising. Brands pay Meta to put products in front of people. That's the entire business.

Now Meta has built a tool that recommends products to you in a conversational format. The product catalog it pulls from is the same pool of brands already paying Meta for placement. The AI layer is new. The business model underneath it isn't.

When an ad company builds a recommendation engine, the recommendations serve the ad business. There's no version of this where Meta's shopping assistant consistently recommends a product from a brand that doesn't advertise on Meta over one that does. That's not a conspiracy — it's an income statement.

Discovery Was Never the Hard Part

Finding products has never been easier. You can search Amazon, scroll Instagram, ask ChatGPT, or now chat with Meta's AI. Discovery is a solved problem.

The hard part is knowing whether what you found is actually good. Are the reviews real? Is this product ranked highly because it's the best option, or because the brand spent the most on ads? Is there a better alternative from a smaller brand that doesn't have the marketing budget to show up?

Meta's tool handles the easy part — finding stuff — and completely ignores the hard part. There's no independent verification layer. No quality signal beyond what sellers and advertisers provide. The AI makes it feel like you're getting a curated recommendation from a smart assistant. You're getting a targeted ad in a new wrapper.

Every Tech Giant Is Running the Same Play

Meta isn't doing anything new here. They're following a playbook that's already proven.

Amazon's Rufus AI drove $12 billion in sales in 2025 and started running ads inside its chat responses in September. Google's AI Overviews increasingly surface Shopping ads. ChatGPT launched a shopping feature backed by affiliate and merchant partnerships.

Every major tech company is shipping AI shopping tools. Every one of them is funded by sellers, not buyers. The consumer's interest — finding the genuinely best product regardless of who's paying for placement — is structurally unrepresented in all of them.

The pitch is always the same: "We're making shopping easier." And they are. They're making it very easy to buy what advertisers want you to buy.

The Data Problem Under the Surface

Even if you set aside the advertising incentive, there's a data quality problem. These AI tools pull from product listings, reviews, and descriptions provided by sellers. That information is unverified.

We already know the review layer is compromised. About 42% of reviews on major marketplaces show signs of manipulation — incentivized reviews, review farms, coordinated posting. When an AI shopping assistant ingests that data as signal, the output is only as honest as the input. Garbage in, confident-sounding garbage out.

Meta's assistant doesn't solve this either. It adds a conversational interface on top of the same unverified product data that already exists. The UX improves. The underlying information doesn't.

What Would Actually Help

A shopping tool that works for buyers would look different from what any of these companies are building. No ad model funding the recommendations. No affiliate commissions shaping which products surface. No seller-provided data treated as ground truth.

That's what we're building at AgentBuy — recommendations based on verified product data and independent sources, with no one paying to influence what shows up.

Meta's shopping assistant will probably work fine for discovering products you didn't know about. Just remember who it's actually working for.