ChatGPT Ads Product Feeds: Catalog Campaigns From Upload to Delta Updates
What a Product Feed Campaign Is
ChatGPT Ads product feeds connect your merchant catalog - titles, descriptions, prices, availability, images, and destination URLs - directly to campaigns, keeping every detail current. Instead of hand-building an ad for every SKU, you connect the feed to a campaign and let OpenAI select an eligible product when the ad serves. OpenAI's own assessment of the format is blunt: ads created from product feeds have been among the strongest-performing ads in its program, which is why feed campaigns are the default recommendation for retail advertisers with broad or frequently changing catalogs. The format trades manual creative control for scale - the catalog's data quality becomes your creative quality.
The Four Layers: Feed, Campaign, Product Set, Template
OpenAI's documentation splits the workflow into four parts with clean responsibilities. The product feed supplies the merchant's current catalog. The campaign carries the familiar controls - budget, schedule, targeting - plus a product_feed mode that binds it to feed delivery. A product set selects one linked feed and optionally filters which products can serve; the documented example limits a shoe catalog to a single brand. A product-ad template defines how values from whichever product gets selected appear in the ad - title, description, and price slot in from the product data at serving time. Understanding the stack from the ground up explains most feed-campaign behavior: diagnosing a weird ad means tracing it all the way back down to the catalog row, not to the campaign.
Uploading the Catalog: CSV, Hosted URL, or SFTP
Feed setup happens in the Feeds area of Ads Manager: create the feed connection, configure its credentials, and get the catalog in through one of three routes. Upload CSV or TXT fits getting started and occasional manual updates. A hosted URL points OpenAI at a stable HTTPS address where your catalog already lives. An SFTP connection serves technical teams that automate server-to-server uploads - OpenAI shows the SFTP location for the account, and the catalog lands there. The file format follows OpenAI's commerce feed specification, published in the developer docs. One structural limit matters for planners: the public Advertiser API provides no endpoints to create a feed connection, list linked feeds, or upload a catalog file - that management lives in Ads Manager, not in the automation API.
Keeping Prices and Availability Current With Delta Feeds
After the initial catalog upload, prices and stock change constantly, and re-uploading the full catalog for every price edit is not the workflow OpenAI built. The Delta Feeds API updates availability, titles, and prices for existing variants directly - no full re-upload. Two boundary conditions shape its use. Access is enabled per ad account: a request without it fails with HTTP 403 and product_feed_api_disabled or product_feed_delta_api_disabled, which is the cue to contact your OpenAI account team rather than retry - the code also separates this access problem from payload validation errors. And the API updates existing feed variants only - it does not create feeds, upload full catalogs, or add products that are not already in the feed, so genuinely new SKUs still ride the catalog path from the previous section.
Beta Boundaries: Ads Only, Not Organic
OpenAI draws one boundary explicitly, and it changes expectations for merchants. Products from a feed are eligible for use in ads; they will not appear in organic ChatGPT conversations, though the help documentation notes that capability may be offered in the future. Product ads themselves use formats designed for showcasing catalog items - product images, titles, star ratings, prices, sale prices, and the brand. For measurement, the same catalog structure carries through reporting: the Ads API supports product-level segments such as product impressions, clicks, and spend, on top of the standard event pipeline where order_created remains the purchase signal.
Wire Feed Campaigns Into Measurement
Catalog-scale delivery raises the stakes on measurement, because every product's bid learns from the same conversion stream. The conversion tracking setup covers the pixel plus conversions API architecture with event_id deduplication, and purchase events should carry the item detail the platform's product shapes expect. Teams coming from Shopify's native stacks can see what changes when measurement moves server-first in the Shopify GTM comparison. MOST delivers tracked conversions server-side to ChatGPT Ads alongside Meta, TikTok, Reddit, Snapchat, and Pinterest, and the free Pixel Activator validates a single event at pixel.way2.us before catalog spend scales.