How to Get Your Products in ChatGPT Shopping

There are four routes into ChatGPT shopping, and most stores need more than one of them.

Route 1: crawlable product pages. ChatGPT search uses a crawler called OAI-SearchBot. OpenAI’s crawler documentation is blunt about it: sites that opt out of OAI-SearchBot will not be shown in ChatGPT search answers. If your robots.txt, your CDN or your firewall blocks it, nothing else in this guide will help much.

Route 2: Shopify Catalog. If you sell on Shopify, OpenAI’s help center says your products are integrated through Shopify Catalog with no extra work. Shopify turns its agentic storefronts on by default for eligible stores. The same is true for Etsy sellers, according to OpenAI’s merchant page.

Route 3: a ChatGPT product feed. Everyone else can apply to share a structured product feed directly with OpenAI. It’s an application, not a self-serve upload yet, and there’s a waitlist. Once you’re approved, you upload a full snapshot of your catalog over SFTP at least once a day.

Route 4: third-party product data. OpenAI says product results draw on structured metadata from first-party and third-party providers. In plain words, ChatGPT also learns about your products from the places that already list them: retailers, marketplaces and product data providers. Studies show a strong overlap with Google Shopping, which we cover later.

Getting into the data is only half of it. Being picked is the other half, and that’s where the AEO work comes in: clear product facts, matching data everywhere, and other websites saying good things about you.

If you’re new to the idea of optimizing for AI answers, our guide on what answer engine optimization is explains the basics. This guide is the ecommerce version.

ChatGPT shopping works in five steps: it detects shopping intent, searches for products, gathers product data, picks a short list, and shows it with links to buy.

Let’s walk through it with an example we’ll use for the rest of this guide. Imagine a hypothetical Bangkok skincare brand, Siam Glow, that sells a mineral sunscreen online to customers in Thailand and the US. A shopper in Los Angeles asks ChatGPT: “What’s a good reef-safe mineral sunscreen for oily skin under $30?”

Step 1: shopping intent. ChatGPT recognizes this as a product question. OpenAI’s help center says that when a question shows shopping intent, ChatGPT shows product options with images, details and links to merchant sites.

Step 2: a shopping search. Behind the scenes, ChatGPT rewrites the question into one or more search queries. Researchers call these fan-out queries. A Peec AI study of 1.1 million shopping fan-outs found shopping prompts produced 1.16 fan-outs per prompt on average, compared with 2.4 for normal search prompts, and the shopping queries were short: about seven words, versus 12 for normal fan-outs (Search Engine Land). Something like “mineral sunscreen oily skin reef safe”.

Step 3: gather candidates. ChatGPT pulls structured product data from first-party providers (feeds merchants send directly, including through Shopify Catalog) and third-party providers (product data services and other sites that list the product).

Step 4: pick a few. OpenAI says not all matching products are shown. The ones that are depend on relevance, the user’s context, price and reviews.

Step 5: show and send. The shopper sees product cards, can open a product for more detail, and clicks a merchant link to buy. For most stores today, the purchase happens on the merchant’s own website.

Shopping research: the deeper mode

There’s a second, slower mode. On 24 November 2025, OpenAI launched shopping research, which asks clarifying questions first, then researches across the web for a few minutes and builds a personalized buyer’s guide with top products, key differences and trade-offs. Search Engine Roundtable reported it runs on a version of GPT-5 mini trained for shopping tasks and is available to logged-in Free, Go, Plus and Pro users (Search Engine Roundtable).

OpenAI’s own video shows the updated experience.

OpenAI · YouTube

OpenAI · An updated shopping experience in ChatGPT · watch on YouTube

If you want the bigger picture of how AI search tools retrieve and rank pages in general, our explainer on how AI search engines work covers retrieval, fan-out and citation in more depth.

ChatGPT picks products on relevance first, then on practical factors like price, reviews, availability and who sells it. OpenAI’s help center spells out more of this than most people realize, so it’s worth reading closely.

What OpenAI says about product selection

According to OpenAI, ChatGPT surfaces items it judges relevant to the question and the user’s context. That context can include Memory and custom instructions. So a shopper who has told ChatGPT they have sensitive skin may see different sunscreens from a shopper who hasn’t.

The factors OpenAI names include price, reviews and ease of use. The inputs include structured metadata from first-party and third-party providers, the model’s own earlier responses, and OpenAI’s safety and product policies.

Two other details matter for merchants. First, ChatGPT may simplify product titles and descriptions, so your exact wording won’t always appear. Second, labels like “Budget-friendly” are generated by the model. They aren’t guarantees, and you can’t buy them.

How ChatGPT picks which store to send the shopper to

Many products are sold by more than one store. For those, OpenAI says the merchant list is built from merchant and product metadata, and ranking considers availability, price, quality and whether the merchant is the maker or the primary seller of the product.

That last factor is good news for brands that sell direct. If Siam Glow sells its own sunscreen on its own site and through two marketplaces, the brand’s own store has a natural claim to the top merchant slot, as long as the price and stock data are accurate.

OpenAI also says the first price shown usually comes from the first listed merchant, which may not be the cheapest. A “Best price” label marks the lowest known price when several merchants sell the item. And price and shipping updates may lag, which is one more reason to keep your feed fresh.

Reviews are summarized, not verified

ChatGPT writes review summaries from public reviews on the web. OpenAI says plainly that reviews and ratings are not verified by OpenAI. In practice that means the volume and tone of your public reviews shape how ChatGPT describes your product, whether those reviews sit on your own site, on a marketplace or on a review platform.

Instant Checkout still exists in a narrow form, but OpenAI has moved away from it. For most stores, ChatGPT now sends shoppers to your site to pay.

September 2025: launch. On 29 September 2025, OpenAI launched Instant Checkout for US users on Free, Plus and Pro, starting with single-item orders from Etsy sellers, with Shopify stores planned next. It ran on the new Agentic Commerce Protocol, built with Stripe and open-sourced the same day (SiliconANGLE).

March 2026: the pivot. An OpenAI spokesperson told Digital Commerce 360: “We’re prioritizing making ChatGPT search and product discovery great,” and “Instant Checkout is moving to Apps, where purchases can happen more seamlessly” (Digital Commerce 360).

Why the change? Shoppers researched heavily in ChatGPT but rarely bought there. Retail TouchPoints reported that Walmart’s Daniel Danker told Wired that ChatGPT’s integrated checkout performed three times worse than sending users to Walmart’s website. The same piece cited a Radial survey of 1,000 US consumers in which nearly two-thirds were uncomfortable sharing payment details with AI agents (Retail TouchPoints).

Late March 2026: discovery first. OpenAI rolled out visual shopping to all users and widened its commerce protocol to cover product discovery. Big retailers including Target, Sephora, Nordstrom, Lowe’s, Best Buy, The Home Depot and Wayfair now send catalog data, and checkout happens in each retailer’s own flow (The Decoder).

OpenAI’s help center still says Instant Checkout is available for some eligible products and merchants. Shopify’s documentation describes the ChatGPT flow as buying on your online store checkout, inside ChatGPT’s in-app browser or a new tab on the web (Shopify Help Center).

For you, this is mostly good news. You keep the customer relationship, the checkout you’ve already tested, your upsells and your email capture. The job is to win the recommendation and make sure the click lands on a page that converts.

The Agentic Commerce Protocol, or ACP, is the open standard OpenAI uses to connect ChatGPT with merchants. Stripe and OpenAI developed it, and it’s open source under the Apache 2.0 license (Agentic Commerce Protocol).

OpenAI’s developer docs split ACP into three flows (OpenAI key concepts).

Product feed. A regularly refreshed file of your products with identifiers, descriptions, prices, stock, images and fulfillment details. This is what makes your products discoverable.

Agentic checkout. A set of endpoints your store exposes so ChatGPT can create and update a checkout session. Your systems still validate the order, calculate tax, charge the card and accept or decline the order.

Delegated payment. A way for OpenAI to pass a one-time, limited payment token to your payment provider. Stripe’s Shared Payment Token was the first implementation.

One line in the docs is worth remembering: OpenAI is not the merchant of record. The store selling the goods is, and it handles refunds and chargebacks as with any other order.

Since the March 2026 pivot, the feed is the part most merchants need. OpenAI’s get-started guide now says to start your ACP integration by sharing a structured product feed (OpenAI get started).

Google has its own version, the Universal Commerce Protocol, and Microsoft supports it in Copilot. We compare the three later in this guide.

Every row in a ChatGPT product feed needs nine fields. OpenAI’s products file reference calls them the basic product data required for a useful discovery feed (OpenAI products reference).

A few rules trip people up. The price format is strict: 24.00 USD, with a space and a currency code, no thousands separators. An empty or unrecognized availability value rejects the whole row, so use unknown explicitly if you don’t know. And don’t use placeholder text such as “n/a” or “null” in optional fields; leave them empty instead.

Optional fields that are worth adding

Variants. If a product comes in sizes or shades, send one row per variant with its own item_id, a shared group_id, listing_has_variations set to true, and a variant_dict such as color and size. Each variant row carries its own price, stock, URL and image.

Identifiers. A valid gtin (8, 12, 13 or 14 digits) or an mpn with the brand helps ChatGPT match your product to the same item elsewhere. OpenAI warns not to invent an MPN to fill a missing GTIN.

Product attributes. Fields such as material, color, size, product_category and dimensions describe the item in the terms shoppers use. For Siam Glow, the category path might be “Health & Beauty > Skin Care > Sunscreen”.

Shipping and returns. shipping_price, accepts_returns, return_deadline_in_days and a return_policy URL. Shipping and returns are exactly the questions shoppers ask before they buy.

Reviews. review_count and star_rating on a 0 to 5 scale, describing the same set of product reviews. Store reviews are a separate field that needs setup during onboarding.

Here’s what a minimal row looks like in JSONL, one product per line:

{"item_id":"SG-MIN-SPF50-50ML","title":"Mineral sunscreen SPF 50, 50 mL","description":"Zinc oxide sunscreen with a matte finish for oily skin.","url":"https://example.com/products/mineral-spf50","brand":"Siam Glow","seller_name":"Siam Glow Store","image_url":"https://example.com/img/spf50.jpg","price":"24.00 USD","availability":"in_stock"}

If you already maintain a Google Merchant Center feed, much of this maps across. Semrush notes that Google-compatible feeds may be accepted if OpenAI confirms it during onboarding (Semrush). Don’t assume it, though. Map your fields to OpenAI’s names and test a sample first.

You submit a ChatGPT product feed in two stages: you apply for access, and once approved you push your catalog to OpenAI over SFTP on a daily schedule.

Stage 1: apply

OpenAI’s docs say onboarding product feeds is currently available to approved partners, and you apply through a form at chatgpt.com/merchants (OpenAI get started).

The form asks for your contact details, company name, headquarters country, website, primary product category, whether your feed meets OpenAI’s spec, and your number of unique SKUs. The page also says that if you’ve already applied, you’re on the waitlist, and that OpenAI is building a self-serve merchant platform (OpenAI merchant page).

Check the prohibited products policy first: OpenAI doesn’t allow adult content, age-restricted products such as alcohol, nicotine and gambling, weapons, prescription-only medicines, unlicensed financial products and other restricted goods.

Stage 2: build and upload the feed

OpenAI’s file upload guide sets out the delivery rules (OpenAI file upload overview).

Format. Parquet is preferred, ideally with zstd compression. Gzipped JSONL, CSV and TSV are also supported. Use UTF-8.

Delivery. Push the file to OpenAI over SFTP. Keep the same file name every time and overwrite it, rather than creating a new file for each run.

Cadence. Upload a full snapshot of your catalog at least once a day. A snapshot is the whole catalog, not just what changed.

Removals. A product missing from a snapshot isn’t deleted right away. OpenAI keeps its last record for up to 14 days. To pull a product from search on the next run, set is_eligible_search to false.

Validate before you scale

OpenAI recommends starting with a small sample of around 100 items with every required field filled, running QA on the first full snapshot, and only then automating. The most common ingestion failures it lists are missing required fields, outdated field names and malformed values.

We’d add one check of our own. Pick ten random rows and open each URL in a browser. Does the price on the page match the feed? Is the variant in the URL the one in the row? Does the image show the right color? Mismatches between feed and page are the quiet errors that cost you trust, because ChatGPT, and the shopper, will see both.

Business Hub is OpenAI’s new dashboard for businesses in ChatGPT, launched in beta on 10 November 2026. If you’re eligible, it’s worth claiming, mainly for the data and the control over your profile.

Here’s what it does, based on launch coverage from VentureBeat and Unite.AI, both published on 10 November 2026 (VentureBeat, Unite.AI).

Claim your profile. Eligible businesses can claim their ChatGPT profile and update the information customers see. Verification can use a business phone number, email, website domain or Shopify account. The person managing the listing then completes a separate identity check through Persona, and a blue check appears once that’s approved.

Connect your products. Retailers can upload a product feed with prices and stock, or connect a Shopify store to sync those details. A Shopify store isn’t required to claim a business.

See Insights. The dashboard shows aggregated activity from the past 28 days: mentions, website clicks and recurring topics in conversations where your website is cited. OpenAI only shows a metric once it passes minimum thresholds, so a blank number may just mean there isn’t enough data yet. OpenAI says it doesn’t share users’ conversations with businesses.

Publish a Business Agent. Eligible US ecommerce and online service businesses can publish an agent that answers shopper questions about products and policies, using information the business provides. Those chats are labeled so shoppers know they’ve moved from ChatGPT to the seller’s agent. Claiming and publishing are free at launch, according to VentureBeat’s report.

Should you do it?

If you’re an eligible US ecommerce business, yes. The 28-day counts are first-party data from OpenAI about how often ChatGPT talks about you, and a verified profile is cheap insurance against outdated facts.

If you’re outside the US, you can still apply to claim a profile where ChatGPT operates and OpenAI supports verification, but publishing a Business Agent was limited to eligible US companies at launch. Keep an eye on it, because OpenAI may widen eligibility over time.

ChatGPT can only recommend what it can read. Most of this is classic technical SEO, with a few AI-specific twists.

1. Let OpenAI’s search crawler in

OpenAI runs several crawlers, and they do different jobs (OpenAI crawlers). OAI-SearchBot finds pages for ChatGPT search, and sites that block it won’t appear in ChatGPT search answers. ChatGPT-User visits a page when a user’s question needs it. GPTBot collects data that may be used for model training, and it’s controlled separately.

That separation is useful. You can block GPTBot if you don’t want your content used for training, and still allow OAI-SearchBot so your products can be found. OpenAI says robots.txt changes can take about 24 hours to apply to search.

Check your CDN and firewall too. Bot protection settings sometimes block AI crawlers even when robots.txt allows them. Our free AI crawler checker shows which AI bots your site lets in, and our guide on whether to block AI crawlers walks through the trade-offs.

2. Put price and stock in the HTML

Vercel’s analysis of AI crawler traffic found that none of the major AI crawlers, including OAI-SearchBot, ChatGPT-User and GPTBot, rendered JavaScript at the time (Vercel, December 2024). If your theme loads the price, the stock status or the variant picker with JavaScript after the page arrives, the crawler may see an empty space where the price should be.

The fix is server-side rendering for the facts that matter: product name in the H1, price, availability, the main specs (as text and HTML tables, never only inside images) and the review summary. Semrush’s November 2026 guide gives the same advice: render price, stock and key specs on the server so they appear in the raw HTML (Semrush).

3. Add Product schema that matches

Product structured data with name, image, brand, an Offer with price and availability, and aggregate ratings helps machines read the page. For products with variants, use ProductGroup. The most important rule is consistency: the schema, the visible page and your feed should all say the same thing. Our guide to schema markup for AEO covers which types help, and the free schema generator builds the JSON-LD for you.

4. Keep facts identical everywhere

Your product title, price, size and ingredients should match on your site, in your feed, in Google Merchant Center, on marketplaces and on any retailer pages. When ChatGPT finds three different prices for the same sunscreen, it has to pick one, and it may not pick yours. If ChatGPT can’t see your site at all, our guide on why a website isn’t showing in ChatGPT covers the full diagnosis.

ChatGPT picks products it can match to a specific question. The more specific facts your product content states, the more questions it can match.

Answer the questions shoppers actually ask

Remember the Siam Glow shopper: reef-safe, mineral, oily skin, under $30. Every one of those is a filter. If the product page says “mineral” but never says “oily skin”, ChatGPT has to infer it, and it may not. If the price isn’t on the page, it can’t confirm the budget.

So start with the questions. Read your customer emails, your chat logs and your reviews. Look at the questions people ask on Reddit or, in Thailand, on Pantip. Then make sure each product page answers them in plain sentences. Our guide to prompt research shows how to build that question list.

Lead with the facts, keep the brand voice

You don’t have to strip your brand voice out. Put the facts first and the story second. A good pattern is a factual first sentence (what it is, who it’s for, the key spec), then a short paragraph on why it’s different, then a spec table, then the brand story. Be honest about limits too: “not suitable for very dry skin” helps ChatGPT match the product to the right person. This is the product-page version of answer-first content. OpenAI’s own feed best practices say the same thing about feed descriptions: concise, factual copy that helps users understand the product (OpenAI best practices).

Reviews and mentions do the persuading

OpenAI says review summaries are generated from public reviews on the web. That means the reviews themselves are a big part of your product content. A product with 400 detailed reviews mentioning “no white cast” and “works under makeup” gives ChatGPT a lot to say. A product with 12 one-word reviews gives it almost nothing.

Ask happy customers for detailed reviews, and make the review text visible on the page as text. Collect reviews where shoppers already look: your own site, marketplaces, Google, and the review platforms in your category.

Then think beyond your own pages. Shopping research reads “quality sources” across the web, and best-of lists, independent reviews, YouTube reviews and forum threads all feed what ChatGPT knows about a product. Digital PR that gets your product into a credible comparison article is AEO work, not just brand work. Our guide on earning AI citations and mentions covers how to build that presence without spam.

ChatGPT isn’t the only AI shopping assistant, and the good news is that the same groundwork serves all of them.

Google

As of November 2025, Google’s Shopping Graph held more than 50 billion product listings, 2 billion of which are updated every hour. In November 2025, Google started rolling out agentic checkout in AI Mode from eligible US merchants, including Wayfair, Chewy, Quince and select Shopify merchants (Google).

Google’s Merchant Center help describes the Universal Commerce Protocol as an open standard for agentic commerce. Merchants can add a checkout button to eligible listings in AI Mode and Gemini, for products eligible in the US, Canada or Australia, and access is limited to select merchants for now. The merchant stays the seller of record (Google Merchant Center Help).

Microsoft

Microsoft Advertising’s August 2026 guidance tells retailers to build an AI-ready product feed in Microsoft Merchant Center so agents can identify and trust what they sell, and describes Copilot Checkout, which lets shoppers buy without leaving the conversation, plus Brand Agents for retailers’ own sites (Microsoft Advertising).

What carries across all three

The common thread is product data. A complete, accurate catalog with stable IDs, GTINs, current prices and stock, honest descriptions and real reviews feeds every one of these systems. Build it once, map it to each format, and keep it in sync. The platform-specific parts, such as which protocol to implement for checkout, can come later and only if the numbers justify it.

You measure ChatGPT shopping with three numbers: how much traffic and revenue it sends you, how often your products appear for the questions that matter, and what OpenAI’s own Insights say if you can see them.

1. Referral traffic and revenue

In GA4, filter sessions by source chatgpt.com and look at sessions, conversion rate and revenue for those visits. If you add utm_medium=feed to your feed URLs, as OpenAI’s best practices suggest, you can separate product-card clicks from other ChatGPT referrals. Shopify also shows AI channel attribution on orders.

Compare that traffic to your other channels. Adobe’s 2026 data gives a benchmark for what good looks like across US retail: so far in 2026, AI-referred shoppers spent 59% more time on site, added 28% more items to carts and bounced 33% less than other shoppers (Digital Commerce 360).

2. A fixed prompt set

Pick 20 to 50 shopping questions your customers actually ask and run them in ChatGPT on a schedule. Record whether your product appears, in what position, and which competitors appear instead. Run each prompt more than once, because AI answers vary from run to run.

This is what we built GrixTrack to do. It tracks brand mentions and citations in ChatGPT, Gemini and Google AI Mode, runs each prompt three times to smooth out that variation, and has a free plan for 5 prompts. Our guide on how to measure AI visibility explains the metrics in detail.

3. Business Hub Insights

If you’re eligible, claim your profile and check the 28-day mention, click and topic data. It’s aggregated and only appears above a threshold, but it’s the only view you’ll get from OpenAI’s side.

Your first 30 days

Here’s the order we’d tackle it in, for most stores.

Expect the technical fixes to show quickly and the reputation work to take months. In the 19 months since April 2025, OpenAI launched shopping, launched checkout, pulled back from checkout and opened a business dashboard. Stores with clean data are ready for whatever comes next.

If you’d rather have a team do this with you, our AEO service covers product pages, feeds, AI crawler access and the reputation work that gets products recommended.

1. OpenAI Help Center: Shopping with ChatGPT Search (checked 11 Nov 2026)
2. OpenAI Help Center: Shopping from Shopify merchants in ChatGPT (checked 11 Nov 2026)
3. OpenAI: Apply to share product data with ChatGPT (merchant application)
4. OpenAI for Developers: Commerce, Get started with product feeds
5. OpenAI for Developers: Commerce, Products file reference (feed fields)
6. OpenAI for Developers: Commerce, File upload overview (SFTP, formats, cadence)
7. OpenAI for Developers: Commerce, Key concepts of the Agentic Commerce Protocol
8. OpenAI for Developers: Commerce, Best practices
9. OpenAI for Developers: Ads product feeds
10. OpenAI: Overview of OpenAI crawlers
11. Agentic Commerce Protocol (Stripe and OpenAI, Apache 2.0)
12. Shopify Help Center: Agentic storefronts
13. TechCrunch: OpenAI upgrades ChatGPT search with shopping features (28 Apr 2025)
14. Search Engine Roundtable: ChatGPT Search gains shopping search features, not ads (29 Apr 2025)
15. SiliconANGLE: OpenAI adds Instant Checkout shopping feature to ChatGPT (29 Sep 2025)
16. Search Engine Roundtable: Shopping research assistant in ChatGPT (25 Nov 2025)
17. Digital Commerce 360: OpenAI shifts checkout plans in its agentic commerce strategy (6 Mar 2026)
18. Retail TouchPoints: Google adds agentic shopping features as OpenAI pivots (20 Mar 2026)
19. The Decoder: ChatGPT is turning into a shopping platform, but no checkout (24 Mar 2026)
20. Retail TouchPoints: OpenAI focuses on discovery, Shopify offers Catalog for any brand (24 Mar 2026)
21. TechCrunch: ChatGPT can now virtually try on clothes for you (1 Oct 2026)
22. VentureBeat: OpenAI lets you claim your business in ChatGPT (10 Nov 2026)
23. Unite.AI: OpenAI launches Business Hub and Business Agents in ChatGPT (10 Nov 2026)
24. Semrush: ChatGPT searches Google Shopping, 100-prompt experiment (20 Jan 2026)
25. Search Engine Land: ChatGPT sources 83% of its carousel products from Google Shopping (Peec AI study, 2026)
26. Semrush: How to optimize for ChatGPT shopping (4 Nov 2026)
27. Vercel: The rise of the AI crawler (17 Dec 2024)
28. TechCrunch: AI traffic to US retailers rose 393% in Q1 (Adobe data, 16 Apr 2026)
29. Digital Commerce 360: Adobe AI referral traffic data, July 2026 (19 Aug 2026)
30. Google: Let AI do the hard parts of your holiday shopping (13 Nov 2025)
31. Google Merchant Center Help: About the Universal Commerce Protocol (UCP) and UCP-powered checkout
32. Microsoft Advertising: How businesses win when AI does the shopping (Aug 2026)

Keep reading

• AI platformsHow to Get Cited by ChatGPT
• AEO basicsWhat Is AEO? Answer Engine Optimization Explained
• AI platformsWhy Is My Website Not Showing in ChatGPT?
• Content & technicalSchema Markup for AEO: What Actually Helps AI Search