ChatGPT Shopping for Merchants: How to Get Your Products Recommended
On this page
ChatGPT recommends products using the product data and web pages it can read. For ChatGPT merchants, that means sharing a complete, current product feed, making the deciding facts clear on your own pages and on the retailer and review pages answers draw on, and testing real buying questions to see whether your products appear.
ChatGPT’s shopping features changed a lot in 2026, so guidance written last year is often out of date.
What changed for ChatGPT merchants in 2026
When OpenAI launched Instant Checkout in September 2025, the idea was that shoppers would buy directly inside ChatGPT. In March 2026, several outlets reported that OpenAI was deprioritizing Instant Checkout as a standalone feature and focusing on how people discover and compare products, with merchants able to use their own checkout (AI Business and Storyboard18, 25 March 2026).
OpenAI’s merchant page sets out the current arrangement (chatgpt.com/merchants, read 2026-10-21):
| Topic | What OpenAI says |
|---|---|
| What merchants share | A product feed, so products can appear in ChatGPT’s shopping results |
| Checkout | Shoppers are sent to your site or app by default, with no fees for purchases made there |
| Why the feed matters | Complete, current data with images, pricing and reviews helps products surface with more accurate information |
| The protocol | The Agentic Commerce Protocol (ACP) supports product discovery and recommendations today, and is meant to expand to the full shopping journey |
| Eligibility | Merchants apply to share their feed; Shopify and Etsy catalogs are already integrated with no separate application |
| Where it is live | Shopping is live for ChatGPT users in the U.S., with plans to add merchants and regions |
For most merchants, then, the work is getting recommended when a shopper asks.
How ChatGPT decides which products to recommend
OpenAI says products are tailored to each shopper based on preferences, context and product data, and that richer product data helps products appear with more accurate information (chatgpt.com/merchants, read 2026-10-21). Beyond your feed, answers also draw on what the web says about a product. Our wider studies found that the material behind an answer is often not on the brand’s own site, and that each engine leans on different sources; see what GEO is for a summary.
Your feed is the one input you control directly. Retailer listings, review roundups and comparison articles still shape how your product is described.
What 9 ChatGPT answers showed
In our skincare study we asked ChatGPT three buying questions from Singapore on 19 October 2026, three times each, in a temporary chat: “best face moisturizer”, “best sunscreen” and “best moisturizer for dry sensitive skin”. The full study is in our guide on how to improve brand visibility in AI search. Here we look only at ChatGPT’s nine answers, from a merchant’s point of view.
| What we looked at | What we saw |
|---|---|
| Sources shown | None. No answer showed a clickable source |
| Local tailoring | 6 of 9 answers referred to Singapore’s climate; 3 gave Singapore dollar prices |
| Brand stability | La Roche-Posay appeared in all 9 answers; Anessa and Beauty of Joseon in all 3 sunscreen answers; CeraVe and Avène in all 3 dry-skin answers |
| Product stability | The brand held, but the product changed. For sunscreen, ChatGPT named UVMune 400 Oil Control in one run, Invisible Fluid in another, and both in the third. For dry, sensitive skin it named three different La Roche-Posay creams across three runs |
Source: WaveWise study log, ChatGPT answers only, Singapore, 2026-10-19.
With no sources shown, you cannot see which page led to a recommendation. That makes the inputs you control, your feed and your own product pages, more important, because they are the only ones you can be sure ChatGPT can read.
Your own range can compete with itself. When ChatGPT recommends a brand, which product it attaches can change from one answer to the next. If two of your products sound alike, the answer may pick either. Clear, distinct descriptions of who each product is for help the right one get named for the right question.
Local details also travel into the answer. ChatGPT tailored most answers to where the question was asked, down to local prices, so accurate regional pricing and availability are part of being recommended.
Limits: nine answers, one engine, one day, one location, one category. Shopping results in ChatGPT are described by OpenAI as live in the U.S.; our runs were from Singapore.
A checklist for ChatGPT merchants
| Area | What to do | Why |
|---|---|---|
| Product feed | Apply to share your feed, or confirm your Shopify or Etsy catalog is connected; keep prices, stock and images current | OpenAI says complete, current feeds help products surface accurately |
| Product pages | State who each product is for, with specifics: skin type, use case, size, key ingredients or specs | Answers recommend products for a need; vague pages give them nothing to quote |
| Range clarity | Make similar products easy to tell apart by name and description | Our runs showed the named product within a brand changing between answers |
| Retailer listings | Check that major retailer pages match your own facts and prices | Answers draw on more than your site |
| Reviews and roundups | Find the review articles that already cover your category and check whether you are in them, and described correctly | Third-party pages shape how products are described |
| Regional data | Keep local prices and availability accurate in every market you sell in | Answers were tailored to location |
| Tracking | Ask your real buying questions several times on ChatGPT and other engines, and record which of your products appear | Answers vary between runs |
Beyond ChatGPT: Google and other engines
Google’s AI features are a separate system. Google says Merchant Center feeds and Business Profiles help products and businesses appear in its AI responses and other Search results (Google Search Central, read 2026-10-19). The work overlaps: the same clear, consistent product data serves both. Other engines, such as Perplexity and Gemini, did not show the same pattern as ChatGPT in our studies, so track each separately.
How to check it worked
After you change something, check three things separately: that the change is live, that ChatGPT has the updated data, and whether your products now appear more often across repeated answers. These are the three levels of verification we use. Our guide on how to track brand mentions in AI search explains how many runs to use and what to record.
Where WaveWise fits
WaveWise Merchant starts from the buying question behind why isn’t ChatGPT recommending us? It reads the answers shoppers would see, records which products were named and why, and turns that into one approved change and a re-test.
A measurement describes the answers during the window in which it was taken, and answers vary between runs, so it is a starting point for decisions. WaveWise does not guarantee that your products will be recommended every time.
If you want a baseline on your shoppers’ buying questions, request a Pilot.
FAQ
Is my store eligible for ChatGPT shopping?
According to OpenAI, merchants can apply to share their product feed, and stores selling through Shopify or Etsy are already integrated without a separate application (chatgpt.com/merchants, read 2026-10-21). OpenAI says shopping is live for ChatGPT users in the U.S. and plans to add more merchants and regions.
Does Instant Checkout still exist?
OpenAI’s merchant page now says shoppers are sent to your site or app by default, with no fees for purchases made there (chatgpt.com/merchants, read 2026-10-21). Reports in March 2026 described Instant Checkout as deprioritized in favour of product discovery.
Do I need the Agentic Commerce Protocol?
ACP is the standard behind product feeds in ChatGPT. If you sell through Shopify or Etsy, your catalog is already connected; other merchants apply through OpenAI’s merchant page.
How do I see whether ChatGPT recommends my products?
Ask the questions your shoppers ask, several times, in a fresh chat, from the market you sell in, and record which products are named. In our runs the product named within a brand changed between answers, so a single check can mislead.
Can I pay to be recommended in ChatGPT?
Not inside the answer itself. OpenAI says ads in ChatGPT appear below the response, are labelled, and do not influence the answer (OpenAI Help Center, read 2026-10-19). What you can improve is the data and pages the answer is built from.
Facts checked: 2026-10-21 · Study data collected: 2026-10-19, Singapore