AI Didn't Disrupt Product Search. It Repackaged It.

A new study just confirmed what the infrastructure has been hinting at for months. ChatGPT isn't discovering products through some proprietary AI reasoning engine. For its shopping carousel, it's pulling almost entirely from Google Shopping organic results. Eighty-three percent of carousel products traced back to Google's top 40 organic listings. Bing accounted for 11%. And of those Bing matches, only 70 products across the entire 43,000-product dataset were exclusive to Bing.

That isn't a coincidence. That's architecture.

ChatGPT isn't discovering products. It's retrieving them. And right now, it's retrieving them from Google. That changes everything about how you approach AI commerce visibility.

Before you write this off as an SEO trade story, stay with me. The implications here go well beyond search marketers. If you sell products and you've been watching AI-driven commerce with one eye, this is the data that should make you look up.

What the Study Found

Researchers at Peec AI analyzed over 43,000 ChatGPT carousel products across 10 verticals (apparel, electronics, beauty, home, toys, and more) and matched them against the top 40 organic shopping results from both Google and Bing for the same queries.

They used a layered matching algorithm that accounted for title rewrites, punctuation differences, and word order variations. A match required a score of 0.8 or above, which in practice means the same brand and the same product. This wasn't fuzzy matching. It was conservative.

The numbers tell a clear story:

• 45.8% of ChatGPT carousel products had an exact title match in Google's top 40 organic shopping results

• When strong matches are included, that figure climbs to over 83%

• Bing strong matches came in at just 11%, and only 70 products across the entire dataset existed exclusively in Bing

• 60% of carousel matches came from Google Shopping positions 1 through 10

• Nearly 84% came from the top 20

ChatGPT carousel position correlates directly with Google Shopping rank. The higher you rank in Google Shopping, the earlier you appear in the ChatGPT carousel.

That's the study. Now let's talk about what it actually means.

Why Google Shopping Works the Way It Does

Most people think of Google Shopping as paid advertising. That's understandable because the first thing you see on a Shopping results page is usually a row of sponsored listings.

But underneath those ads is an entirely separate infrastructure: organic Google Shopping, powered by Google Merchant Center product feeds.

Here's how it actually works.

Retailers submit structured product data to Google through Merchant Center. That data includes product titles, descriptions, GTINs (Global Trade Item Numbers), pricing, availability, images, and category taxonomy. Google ingests this data and indexes it as structured product listings, separate from how it indexes web pages for standard organic search.

When someone searches for a product, Google's Shopping algorithm ranks those listings based on a combination of factors:

• How well the product title and attributes match the query

• The quality and completeness of the product feed

• Pricing competitiveness

• Seller reputation signals

• Click and conversion data over time

The key word in all of this is structured. Google Shopping isn't crawling prose on product pages and trying to infer what something is. It's working from a standardized data schema. Title. Category. GTIN. Price. Availability. Brand. This is why product feed quality matters so much. A vague or keyword-stuffed title performs differently than a clean, accurately categorized one. A feed with missing GTINs behaves differently than one with complete product identifiers.

Now apply that understanding to what the study found.

ChatGPT isn't independently reasoning about which products are best suited to a shopping prompt. It's issuing structured shopping queries, retrieving a set of products from what appears to be Google's organic shopping index, and using that as its selection pool. The retrieval query is short and specific, averaging seven words, because it doesn't need to be anything more than a clean product search. It's targeting a structured index, not the open web.

This is fundamentally different from how ChatGPT assembles a written answer. For contextual responses, it issues longer, more varied search fan-outs averaging 12 words each, uses more of them per prompt (2.4 versus 1.16 for shopping), and pulls from narrative web content that it synthesizes. The two processes are architecturally separate.

The products in the carousel and the text in the response are being built from different sources through different pipelines.

Understanding that distinction is the starting point for any serious AI commerce strategy.

Getting cited in a written AI response and getting your products into an AI carousel aren't the same problem. They don't have the same inputs, they don't have the same fix, and conflating them is costing brands real money.

The Two-Pipeline Reality and What It Means for AI Optimization

Here's where most of the current advice on AI search visibility gets it wrong.

The conversation in our industry has collapsed "being visible in AI" into a single goal. Show up when someone asks ChatGPT about your category. That's reasonable as a starting point, but it papers over a critical strategic distinction. Being cited in a written AI response and appearing in an AI product carousel require fundamentally different inputs.

Getting into a written AI response is about content authority. It depends on the quality of your web presence, the credibility of sources that mention you, the clarity and depth of the content you publish, and the structural signals that AI systems use to evaluate trustworthiness. This is the work I do inside the AI Resonance Model™ framework. E-E-A-T, structured data, named entity recognition, citation patterns, and third-party corroboration all factor in. This is the work of building what I call foundational AI visibility: ensuring that when an AI system retrieves context about your category, your brand is present, credible, and well-represented.

Getting into a product carousel is a different problem entirely. Based on what this study shows, it's primarily a Google Shopping problem. Your product feed has to be clean, complete, and well-optimized. Your product titles have to align with how buyers phrase shopping queries. Your organic Shopping ranking has to be strong enough to land in the top 20, and ideally the top 10, for your priority terms.

These aren't interchangeable. A brand could have excellent content authority: strong editorial coverage, well-structured pages, solid E-E-A-T signals, and still be absent from ChatGPT carousels because their product feed is mediocre. The inverse is also true. A retailer with a pristine Merchant Center feed and top-10 Shopping rankings might never get cited in a written AI response because their content footprint is thin.

The optimization work for each pipeline is real, and neither substitutes for the other.

There's also a third layer worth naming. The study's authors note that final carousel selection and ranking likely incorporates contextual signals from the written-response pipeline: product sentiment, review coverage, and brand mentions pulled from those separate search fan-outs.

So while the selection pool appears to come from Google Shopping, the final curation probably rewards brands that also have strong contextual presence. The two pipelines are distinct, but they aren't entirely independent.

The Paid AI Placement Trap

Ad products inside AI interfaces are coming. Some are already here. And the pitch is predictable: pay to appear in front of users who are querying AI systems with high commercial intent. There's a version of that pitch that will make complete sense for some brands at some point.

But here's the problem with buying visibility before fixing the foundation.

If your Google Shopping feed is underperforming, your organic carousel eligibility is already compromised. The data shows that ChatGPT pulls from the top 40 organic results. Ad placements in AI interfaces sit on top of that organic infrastructure, not instead of it. A sponsored placement in a ChatGPT shopping experience that surfaces a product with weak reviews, a thin product page, and a confusing return policy is still going to lose to a well-prepared competitor at conversion.

Paid AI placements will amplify what is already working. They won't fix what is broken.

Paying to appear in an AI interface before your foundation is in place is the same mistake brands made with paid search in 2005. The spend happens. The results don't.

The brands that will win in AI-assisted commerce are the ones building on a complete foundation: strong product feed data, competitive organic Shopping rankings, credible content presence, and genuine customer sentiment signals. Those inputs determine whether AI systems select you, rank you, and ultimately represent you well. Paying to appear in an AI interface before those inputs are in place is spending money to accelerate a problem.

What to Actually Fix

If you sell products and you want to be in ChatGPT carousels, the work starts in Google Merchant Center, not in an AI ad dashboard.

1. Clean up your product feed titles first. 

Your titles need to reflect how buyers actually phrase shopping queries. Not how your marketing team names products internally. Not how your product catalog was organized five years ago. How real buyers describe what they're looking for when they're ready to buy it. The study used title matching as its primary methodology because that's what the retrieval system appears to prioritize. If your titles aren't clean, specific, and query-aligned, you're invisible to the selection pool.

2. Treat GTINs and product attributes as non-negotiable. 

Google uses these identifiers to match your products against its product knowledge graph. Missing GTINs, incorrect brand fields, and incomplete category taxonomy all reduce your eligibility for organic Shopping ranking. This is basic feed hygiene, but it's astonishing how many mid-market retailers have never fully cleaned it up.

3. Target the top 20 for your priority queries, not just the top 40. 

The data is clear. Nearly 84% of carousel products come from positions 1 through 20. If you're sitting at position 35 for your most important shopping terms, you're technically in the eligible pool, but barely. The optimization goal is top 10.

4. Don't neglect the contextual layer. 

Review coverage, editorial mentions, and comparison content about your products are likely influencing final carousel curation even though they live in a different pipeline. A product that ranks in Google Shopping's top 10 and is well-represented in third-party review content is probably outperforming a product that only has Shopping rank on its side.

5. Measure both pipelines separately. 

If you're currently tracking AI visibility as a single metric, you're missing the signal. Carousel presence and citation presence are different outputs. They have different drivers. They require different interventions. Track them separately or you won't know which problem you're actually solving.

The Bottom Line

This study doesn't mean AI commerce is just Google Shopping with a different interface. The mechanics are more complex than that, and OpenAI could change its retrieval architecture at any time. But it does mean that right now, in March 2026, the primary path into ChatGPT's product carousel runs through the same infrastructure that has always driven Google Shopping performance.

That isn't a consolation prize for traditional search marketers. It's a clarifying signal for everyone.

The AI layer is new. The foundation underneath it isn't. Brands that already built that foundation are ahead. Brands that skipped it don't get a shortcut just because the interface changed.

The AI layer is real. The optimization work is new. But the foundation underneath it isn't. Fix the feed. Earn the ranking. Build the content presence. The AI carousel follows from there.

Not Sure Where Your Foundation Stands?

If this post raised more questions than answers, that's a good sign. Most mid-market brands we talk to have gaps in at least one of these pipelines and don't have clear visibility into which one is costing them the most.

We do a straightforward discovery call — no deck, no pitch. Just an honest look at where your Google Shopping presence and AI visibility stand today, and what's worth prioritizing. If there's nothing to fix, we'll tell you that too.

Sources

1. Peec AI — ChatGPT Shopping Carousel Study, March 2026 (43,000+ carousel products across 10 verticals)

2. Google Merchant Center Product Feed Specifications

3. Google Shopping Organic Ranking Factors