First, a quick look into what's actually happening

When a user asks ChatGPT for a product recommendation, it doesn't search its own database. It fires off shopping-specific search queries (called shopping fanout queries) directly to Google Shopping's organic results. The products that come back from those queries populate the carousel. Paid ads are ignored entirely.

Peec AI confirmed this by reverse-engineering ChatGPT's source code across more than one million shopping queries. The finding: 100% of products seen in ChatGPT Shopping can be explained by the top 40 organic results in Google Shopping for the corresponding query.

That's the situation. Your Google Merchant Center feed is your AI storefront — did you know it yet? 

What follows is a three-level framework for making the most of it, presented by Precis SEO & AEO Director Kristina Bergwall at our second Precis Masterclass. The levels are sequential: each one builds on the last, and each one unlocks more of the opportunity.

Before you start: know which game you're playing

The three levels apply to every e-commerce brand, but your goal within them depends on your business model.

If you're a single-brand retailer (meaning you sell your own products, like a DTC fashion brand or a skincare label), your goal is product recommendation dominance. You want ChatGPT to recommend your specific products when a user asks about your category.

If you're a multi-brand retailer (meaning you sell thousands of products from many different brands, like a beauty retailer or a department store), your goal is to be the supplier. It doesn't matter which brand ChatGPT recommends, as long as your store is where the user ends up buying it.

Both goals use the same three levels. What changes is the emphasis at each stage.

Level 1: Feed fundamentals

The situation

Your Google Merchant Center feed is the foundation of everything. Before any optimisation is possible, your products need to be in the system cleanly — correctly ingested, fully populated, and free of data errors that cause products to be disapproved or suppressed.

This sounds basic, but feed health is more commonly neglected than you'd think, especially at scale. A multi-brand retailer managing tens of thousands of SKUs will almost always have gaps: missing GTINs, inconsistent categorisation, incomplete attributes, duplicate listings. Each gap is a product that doesn't exist as far as Google (and therefore ChatGPT) is concerned.

What to do

Three things at this level:

Ingest

Get your full product catalogue into Google Merchant Center in a clean, structured format (XML or CSV). Every product you sell should be in there. If it's not in the feed, it's invisible.

Audit

Run a feed health check against Google Merchant Center's requirements. Look for disapproved products, missing required attributes, low-quality titles and descriptions, and any data inconsistencies that affect how products are categorised or retrieved. Tools like Peec AI's feed health scanner can automate this process.

Connect

Check your feed against Agentic Commerce Protocol (ACP) readiness standards. ACP is an emerging industry standard for how AI agents access and interact with product data. It's not the dominant system today, but the brands preparing for it now will be better positioned as the landscape evolves.

The goal at Level 1 is simple: 100% indexing efficiency and zero data friction.

Why it works

For single-brand retailers, Level 1 ensures every product you make has a chance of appearing. For multi-brand retailers, it ensures you're a viable supplier for every brand ChatGPT might recommend — which is the entire point of your strategy at this stage.

Malte Landwehr at Peec AI tested this with his wife's food brand. One day after connecting the product to Google Merchant Center, it appeared in ChatGPT Shopping with accurate prices, availability and merchant links. The feed is the on-switch.

Level 2: AI feed enrichment

The situation

Most brands run a single shopping feed, optimised for paid performance. Paid feed optimisation has a specific goal: short, punchy titles that match high-intent search queries and drive clicks efficiently. That's the right approach for paid — but it's the wrong approach for organic AI visibility.

ChatGPT doesn't retrieve products the way a paid auction works. It uses shopping fanout queries that are longer, more descriptive, and often introduce terms the user never typed. When someone asks for "a cozy, heavyweight oversized hoodie for working from home," the shopping fanout query might be "heavyweight oversized hoodie premium cotton structured men." If your product title just says "Oversized Hoodie — Black," you're unlikely to be retrieved.

This is the feed conflict: the same feed cannot optimally serve both paid and organic AI purposes. They need different things.

What to do

Create a separate organic supplementary feed. 

Google Merchant Center supports a supplementary feed alongside your main paid feed, with up to 36 additional product attributes. This is where you have the space to be long-tail, attribute-rich, and context-focused — without compromising the paid feed's performance.

Enrich with AI-powered attribute expansion. 

Manually enriching thousands of product listings isn't realistic. AI enrichment tools can analyse your existing product data and images to identify and add attributes you're missing — material, fit, use case, style descriptors — and rewrite titles and descriptions to include the language ChatGPT's fanout queries introduce.

Concretely, this means:

• Rewriting product titles to include the attribute language ChatGPT surfaces in comparison tables (comfort, durability, versatility, material quality)
• Adding descriptive terms from shopping fanout queries that your products genuinely match
• Using image analysis to surface product features that aren't captured in your existing data
• Formatting brand, size, material, gender and age attributes into clean, comparable fields

Use ChatGPT's comparison tables as a content brief. 

When ChatGPT generates a product comparison for your category, look at the attributes it rates products on. Those attributes (comfort, durability, versatility, style, price range) are a direct signal of what ChatGPT considers relevant. If your product descriptions don't address those attributes with evidence, you're at a disadvantage in the re-ranking layer.

One important note: products with fewer than three stars on the attributes ChatGPT evaluates are effectively filtered out of the carousel. Proving product quality in your feed and on your product pages isn't optional — it's a threshold requirement.

Why it works

For single-brand retailers, a rich organic supplementary feed increases the likelihood that your products are retrieved and ranked well across a wider range of prompts — including the long, conversational queries that characterise how people actually use ChatGPT when shopping.

For multi-brand retailers, enriched product data helps your store surface as the most complete and authoritative source for a given product — which influences both whether ChatGPT recommends it from your store and whether the user chooses you over a competitor when they click through.

Level 3: Conversational domination

The situation

Your feed gets your products into the carousel. But ChatGPT does something your feed can't fully control: it also generates written context around the products it recommends — summaries, comparisons, "what to know about these" sections — and that context is drawn from the broader web.

ChatGPT trains on the entire open web. The sentiment, the language, the associations it has with your brand or your products reflect everything it has encountered about you online: reviews, articles, social media discussions, Reddit threads, YouTube videos, editorial listicles. Your feed tells ChatGPT what your product is. The web tells ChatGPT what people think of it.

This is where the gap between showing up and winning opens up.

What to do

Social SEO

Your brand's presence on Reddit, TikTok, YouTube and other platforms directly shapes the associations ChatGPT has with your products. This isn't just about follower counts — it's about the quality and sentiment of conversations happening about your brand and category. A unified strategy across your marketing, social and content teams for driving brand mentions, product discussions and genuine reviews is the infrastructure that supports long-term AI visibility.

Digital PR and authority listicles. 

41% of commercial AI citations come from listicles — "best [product category] of 2026" style articles on high-authority sites. Being featured positively in those articles influences both the written context ChatGPT generates and, through the re-ranking layer, which products end up in the carousel. This is one of the highest-leverage content investments available right now for e-commerce brands.

For single-brand retailers, this means actively pursuing inclusion in category-level editorial content and managing the review landscape around your specific products. "{Your product} review" SERPs are worth owning.

For multi-brand retailers, the focus shifts to category authority — being the store that editorial content, comparison articles and expert reviews point to as the go-to source for a given category or brand.

Query-expanded content. 

ChatGPT's fanout queries introduce terms that users don't type but that ChatGPT considers relevant. The most common additions in e-commerce include "review," "best," "2026," "vs," and "comparison." These signal the content types that should exist around your products: review content, comparison pages, best-of articles. Creating or earning this content — in English as well as your local market language, since ChatGPT fires English-language fanout queries in 67–95% of non-English searches — closes the gap between your feed and full conversational visibility.

Why it works

LLMs train on the whole web. The brands that control the narrative about their products across the entire digital ecosystem (not just their own channels) are the ones ChatGPT reaches for when forming a recommendation. The feed is necessary. The web presence is what determines whether your brand is trusted, preferred, and consistently surfaced.

As we put it in our masterclass: control the entire narrative.

Where to start

The three levels are sequential, but you don't need to complete one perfectly before moving to the next. Here's a practical way to think about prioritisation:

If you don't have a Google Merchant Center feed yet, start at Level 1. Everything else depends on it.

If you have a feed but haven't audited it recently, audit it first. Feed health issues are invisible until you look for them, and they can silently exclude large portions of your catalogue from both Google Shopping and ChatGPT.

If your feed is healthy and complete, move directly to Level 2. The organic supplementary feed and AI enrichment layer is where most brands currently have the biggest gap — and where the highest-ROI work is available right now.

Level 3 is longer-term and requires cross-team collaboration, but starting to build the social and editorial presence around your products now compounds over time. The brands building it today will be meaningfully harder to displace in AI recommendations six months from now.

The bottom line

ChatGPT Shopping is Google Shopping's organic results, re-ranked by an AI layer, wrapped in a conversational interface that converts 42% better than standard organic traffic. The playbook for winning in it isn't entirely new — it's feed optimisation, product content, brand authority, and digital PR, approached with an understanding of how AI retrieves and ranks products.

If you already have Google Shopping set up for paid, you're 80% of the way there. These three levels close the gap.

👉 Watch the full masterclass: How ChatGPT shops

👉 Want to know where your brand stands in AI search right now? Get in touch with the Precis team.