Products show up in ChatGPT Shopping when OpenAI can read clean, structured data about them: a product feed with accurate identifiers, prices, and live availability, backed by detailed specs, real reviews, and structured data on your product pages. ChatGPT ranks those results by relevance to the shopper's request, not by ad spend, and OpenAI has said the shopping results are organic with no paid placement. One caveat as of July 2026: OpenAI scaled back in-chat Instant Checkout in March 2026 and moved to a discovery-first shopping experience, so the durable work is getting your product data surfaced, not chasing a checkout button. This guide walks through the feed, the attributes, reviews, structured data, and freshness, step by step.
The mechanics here are different from blog-style AEO. Getting an article cited is a job for narrative and clean passages. Getting a product surfaced is a job for machine-readable attributes that an AI shopping agent can compare, filter, and trust in a fraction of a second.
What ChatGPT Shopping and agentic checkout actually are (July 2026)
ChatGPT Shopping is the product discovery experience inside ChatGPT: when a user asks for something to buy, ChatGPT returns a visual set of product results with images, prices, and links, ranked by how well each product matches the request. Agentic checkout is the layer that lets an AI agent complete a purchase on the user's behalf, built on OpenAI's Agentic Commerce Protocol (ACP), which OpenAI co-developed with Stripe and released in September 2025.
The checkout piece has been turbulent. OpenAI launched "Buy it in ChatGPT" Instant Checkout starting with U.S. Etsy sellers and over a million Shopify merchants queued behind them, letting shoppers buy without leaving the chat. As of July 2026, that in-chat checkout has been pulled back: reporting in March 2026 said OpenAI quietly scaled it down after roughly five months, with only about a dozen Shopify merchants live and Walmart reporting that in-chat checkout converted at roughly one-third the rate of sending the same shopper to its own site. OpenAI's public response, announced in late March 2026, was a discovery-first shopping experience with visual results and image-based search, plus dedicated retailer apps (Instacart, Target, and others) rather than a universal in-chat cart.
For merchants, the practical takeaway is stable even though the checkout story is not. Discovery is where the value sits, and discovery is governed by your product data. Whether the final purchase closes in-chat or on your own site, ChatGPT has to surface your product first, and it can only surface what it can read.
Why agentic commerce optimization is about product data, not marketing copy
Agentic commerce ranks structured product data: identifiers, specs, prices, availability, and review scores that an AI agent can parse and compare. It does not reward the persuasive prose that wins on a landing page. An AI shopping agent is not reading your brand story. It is matching a shopper's constraints ("waterproof running shoe under $120, wide fit, size 11") against fields it can filter on, then checking whether independent sources back up the claims.
This is the split that trips up ecommerce teams moving from traditional SEO. On Google, a well-written product page with the right keywords and backlinks can rank. In ChatGPT Shopping, a beautifully written page with a thin or missing feed is invisible, because the agent never had the machine-readable attributes it needed to consider the product at all. The work shifts from writing copy to maintaining accurate, complete, current data. For the broader picture of how AI shopping fits alongside content-driven visibility, see our guide to AEO for ecommerce.
The stakes are worth the plumbing. Salesforce reported AI and agents influenced around $262 billion of the $1.29 trillion in global online spend during the 2025 holiday season, roughly 20 percent, and eMarketer projected AI platforms would account for about 1.5 percent of U.S. retail ecommerce, near $20.9 billion, in 2026. These are correlational market estimates, not guarantees for any single store, but the direction is clear: a growing share of buying decisions now runs through an agent that reads feeds.
Get your products into the feed: the Merchant Program and ACP
The entry ticket to ChatGPT Shopping and agentic checkout is a structured product feed shared with OpenAI through its merchant program, either directly or through a commerce platform like Shopify. OpenAI's ACP documentation describes merchants submitting a product feed (CSV or JSON, with some tooling supporting TSV or XML) containing identifiers, descriptions, pricing, inventory, media, and fulfillment details, starting with a sample feed for validation and then regular snapshots, often daily.
What to do:
- If you are on Shopify, use the native OpenAI/ChatGPT commerce integration to expose your catalog rather than building a feed by hand. Shopify was OpenAI's launch platform partner for agentic checkout, and most merchants will onboard through it.
- If you are not on Shopify, apply to OpenAI's merchant program directly and prepare a feed in a supported format (CSV or JSON) with the required fields for price and availability, plus recommended attributes like media, reviews, and performance signals.
- Keep a clean, current Google Shopping / Merchant Center feed regardless. Product-optimization analyses in 2026 reported that ChatGPT draws a large share of its product data from Google Shopping, so a healthy Merchant Center feed is a second path into AI shopping results even before you are approved for direct integration.
- Expect approval and setup to take time. As of July 2026, direct agentic checkout access is limited to approved partners, and onboarding has been the main bottleneck, not the technology.
Ranking inside the feed is by relevance, not payment. OpenAI has stated the shopping results are not ads and are not influenced by paid placement, so there is no bid to raise. The lever is data quality and match to intent.
Complete every product attribute and identifier
The single highest-impact fix for most catalogs is filling in every product attribute, because missing fields make a product ineligible for the queries it should win. An agent filtering for "size 11, wide, waterproof" cannot return a shoe whose feed omits width or waterproofing, even if the physical product qualifies. Incomplete data is silent disqualification.
What to do:
- Populate a global GTIN (UPC, EAN, or ISBN) for every variant. GTINs let ChatGPT match your listing to the same product described elsewhere and reconcile prices and reviews across sources.
- Fill brand, MPN, category, color, size, material, and every spec a shopper might constrain on. Treat each attribute as a filter someone will apply.
- Keep price and availability accurate to the current state. A feed that shows in-stock when you are sold out gets your product surfaced and then abandoned, which is worse than not appearing.
- Write descriptions for extraction: state the concrete facts (dimensions, compatibility, use case) in plain sentences, not marketing adjectives. The agent is mining facts, not mood.
Build review depth and answer real buyer questions
Reviews and Q&A are a primary trust signal for ChatGPT Shopping, because an AI agent weighs independent validation heavily before recommending a product it cannot physically inspect. A product with 400 reviews averaging 4.6 stars and answered buyer questions gives the agent evidence; a product with three reviews gives it a reason to pick a competitor.
What to do:
- Grow genuine review volume and keep the average honest. Depth and recency both matter; a wall of five-star reviews from two years ago reads as stale.
- Expose review data in your feed and on your product pages so it is machine-readable, not trapped in a JavaScript widget the agent cannot parse.
- Answer buyer questions directly on the product page (fit, compatibility, care, shipping). These map to the follow-up constraints shoppers ask agents, and they give the agent extractable answers.
- Do not fabricate or buy reviews. Agents increasingly cross-check review scores against third-party sources, and a mismatch undermines the trust you are trying to build.
Add structured data to your product pages
Structured data on your product detail pages helps ChatGPT parse and trust the product facts it finds when it visits your site to verify feed claims. Schema.org markup (Product, Offer, AggregateRating) is the baseline format for exposing price, availability, identifiers, and review scores in a machine-readable way that lines up with your feed.
What to do:
- Add
Product,Offer, andAggregateRatingschema to every product page, with the same GTIN, price, and availability values as your feed. Consistency between the two is what builds trust; contradictions erode it. - Do not treat schema as a magic switch. For general AI answer citations, structured data is not required and will not force your brand into a written answer. Its job in agentic commerce is narrower and real: making the product facts on your page unambiguous to an agent that is verifying your feed.
- Make sure the page renders its key facts in HTML, not only through client-side scripts, so an agent that fetches the page can read them.
Keep third-party review and editorial presence
Third-party coverage decides ties, because when two comparable products both match a query, ChatGPT leans on what independent sources say. A product mentioned favorably in respected reviews, buying guides, and community threads accumulates the authority signals an agent uses to choose between near-equal options. This is the same recommendation mechanism that governs how AI search engines evaluate any brand, described in how AI search engines work.
What to do:
- Earn coverage in trade publications and reputable buying guides for the specific use cases your product wins on, not generic roundups.
- Maintain presence on the community and review platforms your category actually lives on. For a deeper channel-by-channel playbook on earning these mentions, see how to get your product recommended by AI.
- Feed the agent a consistent story: the claims in your feed, on your page, and across third-party sources should agree. Agents reward corroboration and penalize contradiction.
Keep feeds and pages fresh
Freshness is a live ranking input in agentic commerce, not a nice-to-have, because an AI agent will not surface a product it suspects is out of stock or mispriced. Stale inventory and price data are the fastest way to get quietly demoted, since surfacing a product that turns out to be unavailable is a bad outcome the agent learns to avoid.
What to do:
- Refresh your feed on a tight cadence. ACP tooling supports frequent updates, in some cases as often as every 15 minutes for price and inventory, and daily snapshots at minimum.
- Reconcile feed, page, and schema values continuously so they never drift apart.
- Update product content when specs, bundles, or variants change, and retire discontinued items promptly rather than leaving dead listings in the feed.
How to know if it is working
The only reliable way to know whether ChatGPT surfaces your products is to ask it the way a shopper would and record what comes back, across engines and over time. Run the buying queries your customers actually type ("best trail running shoes for wide feet under $120"), note whether your product appears, where it ranks, and which competitors and sources ChatGPT leans on. Repeat on a schedule, because AI shopping results vary between runs and shift as feeds and reviews change. Our own experiment in asking ChatGPT where to buy running shoes shows how uneven those results can be.
Manual checks work for a handful of queries. At scale, tracking whether AI search engines name your products across many buying queries is what a monitoring platform automates. The Loudmink AEO platform tracks whether ChatGPT, Gemini, Perplexity, Claude, and Grok mention and recommend your brand and products across the queries buyers use, with plans from $99/mo. It does not manage product feeds or run checkout; it tells you whether the work above is moving your visibility.
Frequently Asked Questions
How do I get my products into ChatGPT shopping?
Share a structured product feed with OpenAI, either through Shopify's native ChatGPT commerce integration or by applying to OpenAI's merchant program directly. The feed needs accurate identifiers (including GTINs), prices, live availability, media, and ideally review data. A clean Google Shopping / Merchant Center feed is a second path in, since ChatGPT draws heavily on Google Shopping product data as of July 2026.
Does ChatGPT Instant Checkout cost money to rank?
No. OpenAI has stated that ChatGPT shopping results are organic and not influenced by ads or paid placement, so there is no bid or fee to rank higher. Products are surfaced by relevance to the shopper's request and the quality and completeness of your product data, not by payment.
What is agentic commerce?
Agentic commerce is buying and selling where an AI agent handles product discovery and, in some cases, checkout on a shopper's behalf. In ChatGPT it runs on OpenAI's Agentic Commerce Protocol, co-developed with Stripe, which defines how merchants share product feeds and how an agent completes a purchase. It optimizes for machine-readable product data rather than persuasive marketing copy.
Do I need Shopify to sell through ChatGPT Shopping?
No, but Shopify is the easiest path as of July 2026. Shopify was OpenAI's launch platform partner for agentic checkout, so its merchants can expose catalogs through a native integration. Merchants on other platforms can apply to OpenAI's program directly with a supported feed, and a well-maintained Google Shopping feed can get products surfaced independently.
Did OpenAI cancel Instant Checkout?
Not entirely, but it scaled it back. Reporting in March 2026 said OpenAI pulled in-chat Instant Checkout after roughly five months due to thin merchant adoption and weak conversion, then shifted to a discovery-first shopping experience with dedicated retailer apps. Product discovery inside ChatGPT continues, so the feed and data work described here still determines whether your products appear.