WooCommerce to a ChatGPT product feed

ChatGPT product discovery reads a feed. It does not crawl your theme and guess. For a WooCommerce store, the practical job is to turn public product records into a file with the nine fields OpenAI currently requires, then keep that file fresh.

What the feed has to contain

Each row is one item or variant. The required fields are item_id, title, description, url, brand, seller_name, image_url, availability, and price. Price is written like 79.99 USD. Availability is in_stock, out_of_stock, pre_order, backorder, or unknown.

Use the SKU you already trust as item_id. If a product has no SKU, assign one in WooCommerce and keep it stable. Do not reuse it for a different item later.

Where the values come from

WooCommerce already stores the title, description, permalink, and images. Brand is often a product attribute or a plugin field. Seller name is your legal business name, not the theme name. Availability should follow stock status. If stock is not tracked, say unknown rather than marking everything in stock.

GTIN is not one of the nine required fields, but Google and Stripe often want a GTIN or an MPN. Copy the number the manufacturer assigned. Leave it blank when you do not have one.

Structured data still matters

A feed helps channels that accept uploads. Crawlers still read the product page. Add Product and Offer JSON-LD to the WooCommerce product template so the price, currency, and availability are in the HTML before JavaScript runs. The visible price and the structured price need to match.

Policies and robots.txt

Link refund, privacy, and terms pages from the footer. Allow OAI-SearchBot and ChatGPT-User on product paths. A block on GPTBot only is a training choice. It is separate from shopping discovery.

What this does not do

Building the file does not enroll the store, and it does not mean ChatGPT will display the products. OpenAI decides what to show. Refresh the snapshot at least daily, and again when a price or stock status changes. CartReady can generate the file from public pages. You still upload it through the path OpenAI documents.