ChatGPT Now Pulls 65% of Product Picks from Structured Feeds — Are You in Them?

Something huge happened in AI shopping this summer, and most e-commerce merchants are completely asleep at the wheel.

Let me lay out three things that happened in the span of a few weeks — three things that, taken together, represent a fundamental shift in how people discover and buy products online.

1. Meta just put an AI shopping agent in WhatsApp

Meta officially launched Muse, a personal AI agent, to US users. It lives inside Meta apps and WhatsApp. And here's the part that matters: it can shop for you.

Not "show you links." Not "suggest some products." Actually execute shopping tasks. Find products, compare options, and purchase on your behalf.

Shopify is a founding partner. When someone asks Muse to find them a new pair of running shoes, the system queries Shopify's product catalog through something called Agentic Storefronts. It matches products automatically. No browsing. No clicking through to your store. The AI just... finds the right product and completes the transaction.

Think about what that means for a second. Your customer isn't even visiting your website anymore. They're telling an AI what they want, and the AI is deciding which store to buy from. Your store either shows up in that AI's product database, or it doesn't exist.

2. ChatGPT Shopping went nuclear

This is the one that really woke me up.

After GPT-5.6 shipped, something dramatic happened to ChatGPT Shopping. The percentage of product recommendations that come from structured product feeds — not scraped web pages, but actual structured data feeds — jumped from 8.26% to 65%.

Let that sink in. Six weeks ago, ChatGPT was mostly crawling the web to find products to recommend, the same way Google does. Now? Two-thirds of its product picks come from structured feeds it directly ingests.

And the fallout was immediate. Out of 687 tracked merchants, 450 experienced significant visibility drops — at least a third of their AI shopping visibility disappeared, almost overnight, around July 10th.

Meanwhile, the top 10 merchants saw their citation share jump from 22.5% to 41.8%. The winners won big. The losers lost big. And about 35% of the feed retrieval is tied to Shopify.

If you're a Shopify merchant, this is your wake-up call. The feed you submit — or don't submit — to these platforms is now the single most important factor in whether AI search engines recommend your products.

3. OpenAI built its own product search engine

Here's the one that got buried in the news cycle but might be the most significant long-term.

OpenAI has built a proprietary search index called Labrador. It's not just one index — it's a family of 12 vertical sub-indexes, and one of them is dedicated entirely to shopping.

Why does this matter? Because right now, OpenAI is A/B testing Labrador to replace traditional web crawling results in shopping scenarios. Instead of scraping your website and trying to figure out what you sell, Labrador uses structured product data from dedicated feeds.

Translation: OpenAI is building its own product knowledge base. It doesn't want to crawl 10 million messy Shopify stores anymore. It wants clean, structured feeds it can query directly.

If your product data isn't in Labrador's shopping index, you're not just missing one AI platform. You're being written out of the foundation that future AI shopping features will be built on.

What all three events have in common

Here's the thread connecting Meta Muse, ChatGPT's feed shift, and Labrador:

AI shopping is moving from "search the web" to "read structured data."

For the last two decades, e-commerce visibility meant Google SEO. Get your pages ranked, get traffic, convert visitors. The rules were messy but the game was the same for everyone.

Now? The game has fundamentally changed.

When someone asks ChatGPT "what's the best wireless earbuds under $100?" — ChatGPT isn't crawling the web anymore. It's pulling from a structured product feed. It's looking at clean data: title, price, rating, availability, specs. It's making a recommendation in seconds, and that recommendation drives a purchase.

When someone tells Meta Muse "I need a new winter jacket" — Muse isn't browsing Shopify stores. It's querying Shopify's catalog directly. Structured data. Clean matching. Instant results.

When Labrador replaces crawling for shopping queries — it's all feeds, all the time. No more scraping. No more guessing. Just structured product data, queried and matched.

The visibility gap is already widening

I've been tracking AI search visibility for Shopify stores every day. The data is unambiguous.

The stores that submitted structured product feeds and optimized their GEO (Generative Engine Optimization) profiles are pulling further and further ahead. They're getting recommended across ChatGPT, Perplexity, and Google AI Overviews simultaneously. They're showing up in multiple AI-driven purchase paths.

The stores that haven't? They're not just invisible — they're becoming more invisible with every update. Because as these platforms shift from web crawling to feed-based retrieval, the stores without structured data aren't just being skipped. They're being excluded.

Think about it like this: two years ago, you could get away with a decent website and good Google rankings. Today, that same website might be completely invisible to AI shopping engines because the data isn't in the right format.

This isn't theoretical. The 450 stores that lost visibility in July — they had decent websites. They probably had decent SEO. But they didn't have their products in the structured feeds that ChatGPT started relying on. And overnight, they lost a third of their AI recommendation visibility.

So what should you actually do?

I'm not here to panic you. I'm here to tell you what I'm seeing and what's actually working.

First: check your visibility. I know this sounds obvious, but most merchants I talk to have never actually tested whether ChatGPT, Perplexity, or Google AI Overviews can find their products. They assume they're visible because they rank on Google. They're not. These are different systems with completely different data requirements.

Second: get your product feeds in order. This is the single highest-leverage move right now. If you're on Shopify, your product data needs to be clean, complete, and available in the formats these AI platforms are ingesting. That means detailed titles, accurate pricing, real availability data, review information, and structured attributes.

Third: optimize for how AI queries products, not how humans search. Humans type "best running shoes for flat feet." AI systems need to match structured attributes: shoe type, support category, foot type compatibility, price range. Your product data needs to speak the AI's language.

Fourth: monitor this weekly, not quarterly. The AI shopping landscape is moving faster than anything we've seen in e-commerce. ChatGPT's feed sourcing went from 8% to 65% in weeks. Meta launched a shopping agent. OpenAI is building proprietary indexes. The rules are changing constantly, and the stores that track their visibility week over week are the ones catching these shifts early.

Why I keep building for this

I built mygeocheck.com because I kept seeing the same thing: merchants spending thousands on ads, thousands on SEO, and zero on figuring out whether AI search engines can even find them.

The tool checks your store's visibility across ChatGPT, Perplexity, Google AI Overviews, and other AI search platforms. It tells you what AI systems can see when they look for products in your category — and what they can't.

With everything that's happened this summer — the feed shift, Meta Muse, Labrador — I've seen a massive spike in merchants checking their visibility for the first time. And the ones who check early, before the next platform update, are always in a better position than the ones who check after they've already lost traffic.

Here's what I know for certain: AI-powered product discovery isn't coming. It's here. The question isn't whether your customers will start using AI to shop — they already are. The question is whether your products are in the databases these systems pull from.

Go check. You might be surprised by what you find — or by what's missing.