ChatGPT ads for ecommerce: product feed campaigns, placements, and store tracking

ChatGPT ads for ecommerce center on product feeds: structured catalogs that OpenAI turns into retail-style ad units and optimizes for clicks or conversions. For a Shopify or other ecommerce store that already runs Google Shopping or Meta catalog ads, the questions are fit, formats, setup effort, targeting, and tracking, with separate organic exposure handled by Shopify Catalog and a broader system covered in this overview of organic product recommendations.

Why ecommerce is a strong documented fit for ChatGPT ads

Per OpenAI's documentation, ads created from product feeds "have been among the strongest-performing ads in our program to date," and the feed system is designed specifically for retail advertisers with broad or frequently changing catalogs. That combination makes ecommerce the clearest documented fit for ChatGPT ads so far, because stores already maintain structured product data that maps cleanly into a feed.

Feed-based campaigns are built to keep up with price changes, inventory shifts, and new products without manual ad editing. For a catalog-driven store, this reduces operational friction compared with single-creative campaigns that must be updated by hand whenever merchandising changes.

A common misconception is that uploading a product feed gets your products into ChatGPT's organic answers. OpenAI's feed documentation states that during the beta, feed products are ads-only and do not appear in organic conversations, while organic visibility for Shopify merchants comes separately through Shopify Catalog, so paid feeds and organic shopping are distinct systems.

Temper And Forge's own testing aligns with this mixed picture. TF has tested ChatGPT ads most heavily on ecommerce, and the campaigns have converted, but consistency has been variable enough that TF cannot extrapolate what performance would look like at full maturity and full spend volume. TF kills ecommerce tests on conversion volume and ROAS from that volume, continues testing ecommerce, and has found that the tracking side exceeded expectations out of the box, including conversion and pixel setup through the OpenAI-built Shopify app.

What ChatGPT product ads look like and display

Product ads from feeds use formats designed for catalog items. Per OpenAI's help materials, these units can show product images, titles, star ratings, prices, sale prices, and brand information.

Because the system is feed-driven, these elements are populated from the fields in your commerce feed rather than from manually written ad copy. That keeps the ad surface consistent with your store and lets OpenAI update prices or sale flags as your feed changes, without you editing creatives.

Feed products themselves are not used for organic answers during the beta period. OpenAI documents that feed products are ads-only, so you should treat the feed as an advertising surface, not as a way to influence unpaid recommendations or rankings.

How product feed setup works and the operating rules

Feed setup starts in the Tools area of OpenAI's ads interface. Per OpenAI's documentation, the flow is: go to the Tools tab, open Feeds, and select Create Feed.

You can then import products in 3 ways: upload a CSV or TXT file manually, provide a hosted HTTPS URL, or connect via SFTP. OpenAI notes that items in a feed expire after 2 weeks, so a one-time CSV upload is not a set-and-forget option. The recommended operating mode is a hosted URL or automated SFTP process that refreshes the feed on a schedule.

Feeds must comply with OpenAI's commerce feed specifications at developers.openai.com/commerce/specs/file-upload/overview. Those specs define the required and optional fields, and OpenAI notes that feed items are ads-eligible by default unless you explicitly mark them as false for ad eligibility.

Once you submit a feed, OpenAI states that processing can take from minutes to hours. The Upload History view shows errors encountered during processing, and the feed summary breaks counts into Ingested, Processed, Not approved, and Ready. These counts describe offer records, not unique products, and the Not approved bucket is an aggregate without per-item reasons in the interface.

Another common misunderstanding is that the Products tab in the ads interface shows your uploaded inventory. OpenAI's documentation clarifies that the Products tab is delivery reporting, not inventory, and it only populates after ads start serving, sometimes taking up to 7 hours after delivery begins. During setup, the reliable way to confirm that products are eligible is to use the ad-group product count after you apply filters.

Targeting and bidding specifics for product-feed campaigns

Product-feed campaigns are created as a distinct campaign type. When you create a new campaign, you select the Product feed type so that the system knows to use your catalog instead of standalone creatives.

Per OpenAI's ads documentation, new product-feed campaigns support geographic targeting at the country level only. This means feed campaigns do not share the full targeting flexibility of other campaign types, and the idea that feed campaigns target like regular campaigns is incorrect in this beta phase. Bulk CSV campaign upload is also not supported for product-feed campaigns.

Within a product-feed campaign, ad groups are where you connect to the feed and narrow which products can serve. You select the feed and then apply product filters, including filters based on ads_metadata custom labels such as custom_label_0 or bidding_tier. OpenAI notes that these labels are used to organize products for ad-group selection and do not impose any geographic restrictions on delivery.

Each ad group has a single ad template. As documented, this template currently supports selecting an image field, and the rest of the ad content is populated from the feed. For bidding, OpenAI states that the Maximize results bid strategy is the default for eligible new ad groups, and this strategy auto-adjusts bids toward clicks or conversions within the campaign budget.

Because the Products tab is delivery reporting that may lag by up to 7 hours after serving starts, OpenAI recommends using the ad-group product count at setup time to verify that your filters are selecting the intended number of products. This is especially relevant when you rely on custom labels to define collections or tiers.

How conversion tracking works for ecommerce and Shopify

OpenAI's measurement stack for ecommerce uses a browser pixel plus a server-side Conversions API. Per the ads developer documentation, these 2 channels use shared event IDs for deduplication so that the same conversion is not counted twice.

Supported commerce events include contents_viewed, items_added, checkout_started, and order_created. Each event can carry ISO 4217 minor-unit amounts and a contents array that includes id, name, quantity, per-unit amount, and a variant_dict object for variant details. This structure lets OpenAI attribute performance down to the product and variant level.

On Shopify, the connection is app-based. As of October 2026, there is an OpenAI-built Shopify app that installs the pixel and Conversions API integration, and there are also third-party apps such as PixelGPT, ChatGPT Ads Pixel, Chatpix, and WeltPixel listed in the Shopify App Store that provide similar pixel and CAPI setups with webhook-based purchase events and deduplicated reporting.

If you need implementation specifics for the pixel, event mapping, or Conversions API payloads beyond this high-level view, those are covered in a separate resource on ChatGPT ads conversion tracking for stores.

What Shopify stores receive organically alongside paid ads

OpenAI documents that Shopify merchants are automatically integrated into ChatGPT's organic shopping results through Shopify Catalog. As of October 2026, this integration does not require any work from the merchant, and organic results are handled separately from ads.

This means that a Shopify store can appear in organic shopping answers via Shopify Catalog while also running paid product-feed campaigns. However, uploading a product feed for ads does not influence or power that organic presence, because OpenAI states that feed products are ads-only during the beta and do not appear in organic conversations.

For a broader view of how organic shopping results work and how they relate to paid investment decisions, see the separate resources on choosing between organic and paid ChatGPT shopping exposure and on campaign setup basics in this guide to starting campaigns in ChatGPT ads.

Temper And Forge has tested ChatGPT ads most heavily on ecommerce. The campaigns have converted, but consistency has been variable enough that TF cannot extrapolate what performance would look like at full maturity and full spend volume, so TF ends ecommerce tests based on conversion volume and ROAS from that volume while continuing to test. TF has found that tracking performance, including conversion capture and the OpenAI-built Shopify app setup, has exceeded expectations out of the box.