Shopping Inside ChatGPT: How to Prepare Your Product Feed
ChatGPT now lets people buy without leaving the chat. Here is what a merchant's product feed needs to contain to be surfaced at all.
rabbitclip teamPublished: 5 min read
Short answer
ChatGPT no longer just suggests products, it can complete the purchase inside its own interface. In OpenAI's own description, this runs on two pieces: a product feed a merchant sends on a regular schedule, and Instant Checkout, a payment flow built on the Agentic Commerce Protocol. A merchant that never sends a feed simply never appears among the products suggested in a chat.
This is not a different SEO discipline so much as a different delivery format. A page gets submitted to Google, a structured file gets submitted to ChatGPT's commerce side; both exist to get the right product in front of the right person at the right moment.
How does product discovery inside ChatGPT actually work?
When someone asks ChatGPT about a product, the search side of the chat scans relevant merchants and listings. What decides inclusion is whether that merchant has already sent a separate product feed to OpenAI; ranking well in classic Google search is not, by itself, enough.
OpenAI's own documentation describes the goal as keeping seller identity, product description, pricing and stock close to real time. Stale or out-of-stock information gets a listing dropped from suggestions.
What does a product feed need to contain?
A product feed is the file that summarises a merchant's catalogue in a format OpenAI can read. It can be sent as CSV, TSV, XML or JSON, and according to OpenAI Developers documentation, updates can be pushed as frequently as every fifteen minutes.
Required fields cover a product identifier, title, description, price, stock status and fulfilment options; extra fields such as imagery, reviews and performance signals are recommended for ranking and trust but not mandatory.
A scene we saw in a home textiles brand's catalogue: titles and prices were tidy, but the size chart and fabric content sat inside a separate PDF catalogue, never making it into the feed itself. The decision criterion here is simple: information living outside the feed's own fields, in another file, never reaches the model at all; however accurate a size chart is, it counts for nothing until it moves into the feed. What a merchant needs to do is fill every field the feed asks for from one single source, once.
- A product identifier (SKU or an equivalent unique code)
- A title and a plain description, stripped of marketing language
- Current price and currency
- Stock status (in stock or not, quantity)
- Delivery timeframe and region
What is Instant Checkout and the Agentic Commerce Protocol?
Instant Checkout is the flow that lets a buyer finish paying without ever leaving ChatGPT. It runs on the Agentic Commerce Protocol, an open standard covering communication between buyer, AI agent and merchant, integration with payment providers such as Stripe, and order notifications.
For a merchant to support this flow, the checkout steps OpenAI defines, buyer details, fulfilment and payment, need to be implemented on the merchant's own systems. That is a software integration, not a one-line setting.
Which platforms and regions are covered?
OpenAI announced Instant Checkout starting with merchants on specific platforms, and coverage by platform and country keeps widening. Whether a merchant's own platform, Shopify or a UK marketplace among them, supports this flow directly depends on the situation; the current list sits on OpenAI's own developer page.
Even without direct platform support, a merchant can still convert its catalogue into the required format manually or through a middle layer; in that case the goal is product visibility rather than the checkout flow itself.
The first step for a store: a practical checklist
The most concrete way to start appearing on ChatGPT's product side is turning the existing catalogue into a single, accurate, regularly refreshed file.
- Check the export options of the current platform (Shopify, WooCommerce, or an equivalent).
- Compare the exported file against OpenAI Developers' product feed schema and fill any missing fields.
- Confirm price and stock fields update automatically rather than being entered by hand.
- Manually check for a week after the first submission that descriptions and prices display correctly.
- Resend the feed on a regular schedule, ideally daily or whenever stock changes.
Common mistakes
A common one we have seen with a workwear manufacturer is sending the feed once and leaving it untouched for months; a sold-out item keeps showing up in suggestions until a shopper clicks through and hits an out-of-stock message.
Another is filling product descriptions with marketing language; words like 'unique' or 'premium quality' do nothing inside a feed, the model is looking for concrete attributes such as material, size and colour.
When does this not apply?
Not every merchant needs to send a separate product feed today. A small craft workshop with fifteen or twenty products, updated by hand, usually gets little return from building one; if customers already search for the brand by name, effort belongs elsewhere first.
The decision criterion comes down to two questions: how large the catalogue is and how often stock actually changes. A merchant with hundreds of SKUs and stock shifting daily gets a fast payoff from a feed, because manual tracking simply becomes impossible; a business with ten or fifteen products and stock that stays put for months gets a far smaller return on the same investment.
If a business has not yet stated the basics on its own site, price, delivery time, return terms, that is where the priority belongs first; a feed is a layer added on top of a foundation that already works, not a shortcut that fixes a foundation that does not.
Appearing inside ChatGPT does not need a complex SEO trick, it needs an accurate feed kept alive on a schedule. For most merchants the real work is getting the catalogue into the right format once and keeping that data current. A short call with rabbitclip is enough to map which flow fits a given platform.
FAQ
Is a platform like Shopify required to show products in ChatGPT?
No. Without direct platform support, product data can still be converted into the required format manually or through a middle layer.
Does Instant Checkout work in every country?
It depends; coverage varies by country and by the merchant's integration, and the current list sits on OpenAI's own page.
How often should a product feed be updated?
As soon as stock or price changes, at least daily; OpenAI allows updates as frequently as every fifteen minutes.
Who builds the feed, the developer team or marketing?
Usually both: the data structure is built by developers, and the accuracy of product descriptions is checked by the content or product team.
