Short answer: ChatGPT Ads product feed campaigns let retail advertisers upload a product feed in Ads Manager and create ads based on the catalog. Products from the feed are eligible for ads during the beta — they do not appear in organic ChatGPT conversations. Feeds require a minimum of 1,000 products and a maximum of 2 million, uploaded via SFTP.
Who should use product feed campaigns?
| Business type | Product feed fit | Reason |
|---|---|---|
| Ecommerce store with 1,000+ SKUs | Strong fit | Meets minimum feed scale and can segment by product line |
| Retailer with frequent inventory changes | Strong fit | Feed keeps product details and availability current |
| Small store with 50 products | Poor fit | Does not meet OpenAI's listed minimum feed size |
| Lead-gen service business | Not a fit | No product catalog |
| Marketplace with millions of products | Potential fit | OpenAI lists a maximum of 2 million products per feed |
What product data matters most?
The most important product data is the information that helps ChatGPT understand which product is relevant to a user's conversation. Product title, description, category, image, landing page, price, availability, and metadata fields need to be accurate and consistent.
| Field | What to check | Bad example | Better example |
|---|---|---|---|
| Product title | Clear product type and distinguishing attributes | "Classic Pro" | "Classic Pro Carry-On Suitcase, 22-inch, Navy" |
| Description | Real use case and product details | "Best bag ever" | "Lightweight carry-on suitcase for 2 to 4 day trips with laptop sleeve" |
| Image | Product is clear and accurate | Lifestyle image where product is tiny | Clean product image with visible details |
| Landing page | Product page matches feed | Redirects to homepage | Exact product URL |
| Availability | In-stock status is current | Shows sold-out item as available | Feed matches actual inventory |
| Metadata | Useful grouping fields | Blank | product_line, bidding_tier, margin_band |
How should ecommerce brands structure ad groups from a feed?
Structure ad groups by product line, use case, margin tier, inventory priority, or buyer intent. OpenAI documentation says advertisers can use product filters to define which products are eligible for an ad group, and can use the ads_metadata field if available filters are not enough.
| Ad group | Product filter logic | Context theme | Why it works |
|---|---|---|---|
| Weekend carry-ons | product_line = carry_on | Short-trip packing and airline carry-on needs | Strong use-case fit |
| Business travel bags | use_case = business_travel | Laptop bags and work travel organization | Clear buyer situation |
| Giftable travel gear | giftable = true | Gift ideas for frequent travelers | Seasonal and conversational |
| High-margin luggage | margin_band = high | Premium buyers comparing durable bags | Better profitability control |
| Clearance travel accessories | inventory_status = clearance | Shoppers looking for deals | Separate economics and messaging |
Product title formula
Brand + product type + key attribute + size/color + use case
| Product | Weak title | Better title |
|---|---|---|
| Carry-on suitcase | "RoadMax" | "RoadMax 22-inch Carry-On Suitcase, Lightweight, Black" |
| Office chair | "Ergo Plus" | "Ergo Plus Adjustable Office Chair with Lumbar Support" |
| Golf travel case | "Tour Guard" | "Tour Guard Hard-Shell Golf Travel Case for Airline Travel" |
| Kids lunchbox | "Lunch Buddy" | "Lunch Buddy Insulated Kids Lunchbox, Leak-Resistant, Blue" |
Product feed launch checklist
- –Feed uploaded through SFTP and processed successfully.
- –Product count matches the intended set.
- –Ad group filters include the correct products.
- –Product titles and descriptions are clear in previews.
- –Images show the correct products.
- –Landing pages resolve and match the product.
- –Out-of-stock products are excluded or marked correctly.
- –Metadata fields are filled for product grouping.
- –Ad template previews correctly.
- –Campaign objective, budget, bid, dates, and geography are correct.
- –Product claims comply with OpenAI policies.
How should ecommerce brands measure product feed campaigns?
| Metric | Why it matters | Source |
|---|---|---|
| Spend | Media cost | Ads Manager |
| Clicks | Traffic volume | Ads Manager |
| Purchases | Attributed transactions | Ads Manager plus ecommerce backend |
| Revenue | Top-line value | Ecommerce platform |
| Gross margin | Real economics | Internal product data |
| Return rate | Net quality | Ecommerce platform |
| Product group ROAS | Segment-level profitability | Combined report |
Common mistakes
- –Uploading a feed with vague titles.
- –Using manufacturer descriptions without useful attributes.
- –Not separating high-margin and low-margin products.
- –Sending product clicks to collection pages when product pages exist.
- –Treating feed eligibility as organic ChatGPT visibility.
- –Ignoring availability and price freshness.
Related ChatGPT Ads guides
- → How to run ChatGPT Ads — setup, campaign structure, and launch checklist
- → ChatGPT Ads conversion tracking — Pixel, Conversions API, and UTM setup
- → ChatGPT Ads ROI — measuring ecommerce performance without mature benchmarks
- → Brand search cannibalization in PMax — how AI campaigns affect ecommerce ROAS