OpenAI Ads for Shopify, WooCommerce and BigCommerce shops

Published 14 September 2026 · Based on the OpenAI Ads API documentation

A shop that wants its products in ChatGPT answers needs four things in place before the first campaign: an ad account, a product feed, a pixel and server-side conversions. This guide explains each one, how Adsonomy sets them up for Shopify, WooCommerce and BigCommerce, and which stock and margin signals then drive the campaigns.

What does a shop need to advertise in ChatGPT?

Four things. An OpenAI Ads account at ads.openai.com with an API key. A product feed in the Agentic Commerce product file format, delivered over SFTP. The OpenAI pixel on the storefront. And server-side conversions through the Conversions API, deduplicated against the pixel so no order is counted twice. Feeds, pixels, Conversions API keys and conversion-optimised bidding are capabilities that OpenAI switches on for an account; they are not on by default.

The ad account is yours. Adsonomy connects to it with an API key you issue in the Ads Manager settings, or manages it under a partner key. Read how ads in ChatGPT work for the account structure and the formats.

How does the product feed work?

The feed is your catalogue in the Agentic Commerce product file schema that OpenAI publishes for commerce. A feed is registered through the API, the file is uploaded over SFTP with a password or an SSH key, and later changes can be sent as deltas rather than full re-uploads. Product ads then read the item's image, URL and price directly from the feed.

A feed-based ad group points at a product set, a subset of the feed, and carries one active product ad template. The template can print the current price, so a price change in the shop shows in the ad without touching the creative. Accounts with the capability enabled can also report which products received zero impressions, which is the quickest way to find items whose hints or data never match a conversation.

The API does not document an endpoint to list existing feeds, so whoever manages the account keeps its own record of feed IDs and uploads. Adsonomy keeps that registry per shop, with the last successful upload and any rows the schema rejected.

How does conversion tracking work for a shop?

Two layers report the same events. The pixel runs in the browser and captures page views, product views, add to cart, checkout started and orders. The Conversions API sends the same events from your server, which is more reliable because it survives ad blockers and abandoned tabs. Both can fire for one order; the platform deduplicates them by event ID, and the pixel's browser reference is passed along with the server event to tie the two together.

Personal data in a server event, such as the customer email, is normalised and hashed with SHA-256 before it is sent, as lowercase hexadecimal. Events must carry a timestamp within the last seven days. Of the fourteen standard events on the platform, the five that matter for a shop are page_viewed, contents_viewed, items_added, checkout_started and order_created. Only standard events can be set as an optimisation goal, and only order_created feeds the attributed sales and ROAS metrics in reporting.

The click attribution window is configured by the advertiser; OpenAI's documentation uses 30 days as its example. Conversions take at least a day to appear in reporting, so daily decisions have to be made on yesterday's orders, not this morning's.

The pixel loads from OpenAI's script domain and needs a pixel ID. If your storefront enforces a content security policy, that domain has to be allowed. Adsonomy checks this during setup and installs the tag where the platform allows it.

How does Adsonomy connect Shopify, WooCommerce and BigCommerce?

With one authorisation per platform. Adsonomy reads the catalogue, stock levels, prices and orders through the platform's API, builds and maintains the feed, installs the pixel, and streams every order to the Conversions API as an order_created event with a hashed customer identifier and an event ID that matches the pixel event.

Shopify

You authorise Adsonomy against your store. Products, variants, inventory levels and orders sync continuously. Variants map to feed items so that a colour or size that sells out is removed from the product set on its own, without pausing the whole product.

WooCommerce

Adsonomy connects through the WooCommerce REST API on your WordPress site and places the pixel through the site's head. Stock management has to be enabled in WooCommerce for stock-out protection to work; without it, Adsonomy can only see price and availability flags.

BigCommerce

Adsonomy connects as a store app and reads the catalogue, inventory and orders through the BigCommerce API. Multi-storefront setups are treated as one catalogue with one feed unless you split them into separate ad accounts.

Shopware, Magento and PrestaShop follow after launch on the same feed, pixel and conversions pipeline. If you sell on another platform, join the waitlist and tell us which one.

What stock and margin signals change?

Stock and margin decide what gets promoted. When a product runs out, its product set and ads are paused before the budget is wasted on a page that cannot sell, and they resume when stock returns. Margin joins the reporting so the goal can be profit per order rather than a ROAS figure that ignores the cost of goods.

The OpenAI platform knows your revenue from order events but nothing about your costs. Adsonomy pulls cost or margin per product from the shop where it exists, or lets you set it per category, and uses it in three places: to pick which products get their own ad groups and hints, to set the ceiling a bid controller may push a fixed bid to, and to report profit next to spend in the daily brief. Price changes flow into the feed automatically, so a promotion shows the promotional price in the ad, and a promo calendar can raise or lower budgets around it. The guardrails guide explains how those changes are limited and logged.

What should EU shops know about custom audiences?

Custom audiences, uploaded lists of customer emails or phone numbers, are not available for campaigns targeting the EEA or Switzerland. This does not hold back a shop: on this platform the targeting is done by context hints and the product feed, not by audiences. Adsonomy was designed without them.

For campaigns outside the EEA the platform accepts audience files in CSV or TXT up to 500 MB, with plain or hashed emails and phone numbers, and forbids data bought from brokers. Adsonomy applies the same rule everywhere: personal data is hashed before it leaves your systems, conversion events carry only what the platform needs, and the service is hosted in Amsterdam.

Frequently asked questions