ChatGPT Ads for e-commerce: three webshops, four countries, €0.44 per click

The order value and the 3.0x ratio follow Ads Manager attribution: 30 days after a click and 1 day after an impression. This is not verified revenue. See How we read the numbers.

For comparison: the average CTR on Meta is 1.49% (Digital Applied, Q1 2026). In ChatGPT, these webshops get a higher click rate at a low cost per click. The markets differ, so read this as an indication.

The company sells replacement parts and accessories. For these products, the use case decides which version fits. A buyer wants to know if a part fits their appliance or model.

A buyer can also ask ChatGPT that question. They know the appliance or the problem and ask for help choosing. That is the opportunity for a specialist webshop: an ad directly below that answer.

The company had three requirements:

• The setup must reuse the existing product range and product feeds.
• The team must manage three webshops and four markets next to the existing Google Ads campaigns.
• Measurement must show orders and product interest separately.

1. Product feeds via Channable

Channable prepares the existing product data. The feed mapping connects the fields from the product range to the fields ChatGPT Ads uses. Ads Manager fetches the three feeds over SFTP.

The product templates use product.title, product.body and product.price. A price change in the webshop reaches the ad with the next feed update.

This makes the product data the ad copy. For parts, model, size and connection matter. A title without that data does not help the buyer, in the ad or on the product page. OpenAI explains how this works in its documentation on product campaigns.

2. Turning the product range into commercial selections

A feed contains all products. The selection per ad group decides which products that group shows.

We used revenue data and bestsellers from the webshops as the starting point. Then we split the range with three filters: category (product_category), brand (brand) and product ID (item_id). Seven product groups use a fixed list of product IDs, because a category is too broad there.

3. A recognisable structure for three webshops

We copied the naming from the Google Ads account. Every campaign name shows webshop, ad format and theme. The structure has three levels: campaign (goal, budget, country), ad group (theme or product selection) and ad or product template.

The three product campaigns show products directly from the feed. The two text campaigns cover themes such as replacement, fit and use, each with its own landing page. All ads carry UTM tags, so web analytics can identify the traffic.

4. Turning product knowledge into context hints

ChatGPT Ads works without keywords. The system picks an ad based on the context of the conversation. Context hints give OpenAI extra information on when an ad group is relevant (OpenAI's explanation).

We wrote the hints as natural descriptions of situations. Hint, ad and landing page always address the same choice question:

These examples illustrate the method. They are not literal account copy. OpenAI does not share user conversations with advertisers (source). So we build the hints from product knowledge and the content of the webshops.

5. Measuring conversions in the browser and on the server

The OpenAI pixel measures in the browser. Conversions API measures from the server. Ads Manager diagnostics show events through both routes for all three webshops. A shared event_id stops the same order from counting twice (OpenAI's explanation).

We gave each campaign type its own measurement goal:

This keeps orders and page views apart in reporting. We calculate cost per conversion per event type, separately for click conversions and for the total including impressions (OpenAI's attribution rules).

6. Management after launch

The campaigns went live in phases from 5 September. During the rollout we added product groups, adjusted selections and lowered bids. In the final week of the measurement window we added ten product groups.

We check every change against three questions:

• Does the product selection match the theme of the group?
• Do the ad and landing page answer the same choice question?
• Does the measured conversion type match how we judge the campaign?

We report what the platform measures and state what that does and does not prove.

• Ads Manager attributes the €3,006 order value. Part of it can come from impressions without a click (view-through). Verified webshop orders, returns and margins are not yet connected.
• OpenAI processes attributed conversions within 24 to 48 hours, and totals can still change after that (OpenAI).
• The measurement window is short and the rollout was phased. The numbers show a first picture, not a stable level.

The next step connects campaigns to verified orders and margins. Then the team can judge each campaign and product selection on real profit.

Ads do not influence ChatGPT's organic answers (OpenAI). Whether ChatGPT itself mentions and correctly describes these webshops, we measure separately with AI Visibility.

• Live early. This client's first ads ran on 5 September 2026, within a week of Ads Manager opening to Dutch advertisers.
• Experience across accounts. We manage ChatGPT Ads for advertisers in the Netherlands, Germany, France and the United States, in e-commerce, retail and B2B.
• OpenAI Select Partner. Read what that means.
• The trainer the market chose. Arjan ter Huurne teaches the ChatGPT Ads courses at Emerce Academy and Beeckestijn Business School.
• Paid and organic together. We also measure what ChatGPT says about a brand organically. The organic answer sits directly above the ad.