How to Get Your Products Recommended by ChatGPT

To get your products into ChatGPT’s shopping results, you need a connected product feed: either through Shopify Catalog if you sell on Shopify, or a feed you submit to OpenAI directly. After that, what ChatGPT says about your product depends on how clear your product data is and what the rest of the web says about you.

That is the short version. The longer version matters because the rules changed twice this year, and a lot of the advice still circulating was written for a system that no longer exists.

What changed in ChatGPT shopping in 2026?

Two things.

First, OpenAI walked away from in-chat checkout. On March 24, 2026, it said the first version of Instant Checkout “did not offer the level of flexibility” it wanted, and moved its focus to product discovery, with shoppers finishing the purchase on the merchant’s own site or app (TechInformed, March 30, 2026). Uptake had been thin. A Forrester analyst put Shopify adoption at about 30 merchants as of February, and Walmart said ChatGPT checkout converted at about one third of its own website, as reported by Stellagent.

Second, feeds became the entry ticket. The Agentic Commerce Protocol (ACP), originally built for checkout, is now the catalog and discovery layer. Merchants send structured feeds with titles, descriptions, images, price and availability, as a daily snapshot plus API updates. Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair have integrated it, and Shopify merchant data flows in through Shopify Catalog. Then, in early September, Search Engine Roundtable reported that ChatGPT shopping results now feature only brands that submit product feeds (September 4, 2026).

So if you have no feed, you are not in the shopping results. It does not matter how good the product is.

How does ChatGPT decide which products to recommend?

OpenAI has not published a ranking formula. What it has said is that shopping results “are not ads and are not influenced by partnerships.” Advertising is a separate, labeled product, which we cover in our post on ChatGPT ads for ecommerce brands.

The most useful outside framework we have seen is from Similarweb, which describes three inputs that need to agree with each other:

1. The feed. ID, title, description, price, availability, images and attributes.
2. Your product pages. Used to check specs, imagery and who the product is for.
3. Your brand footprint. Reviews, marketplace listings, editorial roundups and other trusted sources.

That matches what we see when we test prompts for clients. ChatGPT tends to name products whose facts are easy to confirm from more than one place. A product whose feed says “waterproof”, whose page says “water resistant” and whose reviews complain about leaks gives the model three conflicting answers. It will usually pick someone else.

Similarweb also found that journeys which start in ChatGPT and end in a purchase somewhere else (Google, a retailer’s search box, a marketplace) happen about 5 times more often than journeys with a direct click from ChatGPT to a retailer. Most of ChatGPT’s influence is upstream of the click. We look at what that means for measurement in does AI traffic convert.

What if most of my sales are on Amazon?

This is the awkward part for Amazon-first brands. Amazon has blocked OpenAI’s crawlers in its robots.txt, including ChatGPT-User, the agent ChatGPT uses to fetch live pages when someone asks a question, and OAI-SearchBot (Modern Retail, November 21, 2025). ChatGPT cannot reliably read current Amazon prices, stock or detail pages.

That does not make ChatGPT irrelevant for you. It means your Amazon listing cannot be the only source of truth about your product. You need something ChatGPT can read and send people to: a small DTC site with a feed, even if most buyers end up on Amazon anyway. We cover the wider picture in why ChatGPT rarely sends traffic to Amazon.

The checklist: a brand with an Amazon listing and a small DTC site

This is the order we would work in. Feed first, because nothing else counts until you are eligible.

1. Get a feed connected. If your DTC site runs on Shopify, Agentic Storefronts is on by default for eligible US stores and your products reach ChatGPT through Shopify Catalog. Check it under Sales channels in your admin (our Shopify Agentic Storefronts guide has the steps). If you are on another platform, you can apply for direct feed access at chatgpt.com/merchants, or use Shopify’s Agentic plan to list in Shopify Catalog without moving your store.
2. Fill every required field. One vendor guide to the OpenAI feed spec, from Alhena (updated September 30, 2026), lists nine required fields: item_id, title, description, url, brand, seller_name, image_url, availability and price, with a full snapshot at least daily. Confirm the current spec with OpenAI before you build anything, since this is changing quickly.
3. Write titles and descriptions that answer questions. People ask ChatGPT things like “a carry-on backpack that fits under a plane seat and holds a 16-inch laptop.” Put size, material, compatibility and use case in plain words, not just keywords. This is the same discipline we apply in listing optimization on Amazon.
4. Make the feed, the page and the Amazon listing say the same thing. Same dimensions, same claims, same product name. Contradictions are the most common problem we find.
5. Keep price and stock current. A product that shows as in stock at the wrong price creates a bad experience, and feeds are refreshed daily, not instantly.
6. Add structured data to your product pages. Product, Offer and AggregateRating schema help any machine reader confirm the basics (Yotpo covers this well).
7. Earn mentions outside your own site. Reviews with real detail, inclusion in independent roundups, and honest answers on forums all feed the “brand footprint” input. You cannot buy this quickly, and paying for fake reviews will hurt you on Amazon long before it helps you in ChatGPT.
8. Test monthly. Ask ChatGPT the five or six questions your customers really ask. Record whether you appear, what it says, and which sources it cites.

How do you know if it's working?

Be realistic about what you can measure.

If you only watch the first row you will undercount what ChatGPT is doing, for the reason Similarweb gives above. If you only watch the last row you will chase a number that does not pay the bills. Watch all four for three months before you decide anything.

What we'd do this week

For a typical client of ours, a brand doing most of its volume on Amazon with a Shopify site on the side:

1. Log into Shopify and confirm the Agentic Storefronts channels are on, and that the products you care about are in the catalog.
2. Export the product feed and compare it, line by line, against the Amazon listing for your top five products. Fix every mismatch.
3. Rewrite the first two sentences of each product description so they state what the product is, who it is for and the one spec people ask about most.
4. Write down ten real customer questions and test them in ChatGPT. Save the answers as your baseline.
5. Decide whether paid placement is worth testing. For most brands we would wait until the organic basics are done, for the same reason we treat PPC as an amplifier rather than a fix: ad spend on a weak listing just buys more of the leak.

None of this replaces your Amazon work. It is a few days of product data clean-up that makes your brand easier for every AI assistant to describe accurately, including Google’s, which we cover in selling in Google AI Mode. If you would like a second pair of eyes on your listings and feed before Q4, our free audit is a good place to start.

All rules and requirements here are as of October 2026. Confirm the current feed spec and eligibility with OpenAI and Shopify before you act.