# ChatGPT ads for Amazon sellers: where the next euro goes URL: https://adsonomy.com/guides/chatgpt-ads-for-amazon-sellers Published: 20 September 2026 Last updated: 20 September 2026 Plain-text version of the page above. Adsonomy is an independent product and is not affiliated with OpenAI or Amazon. ## Summary Written for a seller who already pays an Amazon referral fee, usually an FBA fulfilment fee, and then a Sponsored Products bid on the same unit. Sets out the loaded cost of an Amazon unit under the fee schedules checked in May 2026, why defending your own detail page is a standing cost, what a ChatGPT ad euro buys instead, how to choose between your own shop and the Amazon listing as the destination, how to measure each one, and what Sponsored Products still does better. For the two platforms compared as advertising systems, see https://adsonomy.com/guides/chatgpt-ads-vs-amazon-ads. ## Key facts - On amazon.de, .fr, .it, .es, .nl and .com.be the referral fee is 15% for most general merchandise, 13% for DIY & Tools and 7% for Computers, Consumer Electronics, Tyres and Full-Size Appliances, under the schedule effective 5 January 2026 (checked May 2026). - It is charged on the gross, VAT-inclusive price for consumer orders, with a minimum of EUR 0.30 per item on those six marketplaces. Business orders are charged ex-VAT, so the fee lands on the net price. - FBA fulfilment is a separate per-unit fee by size tier and charged weight, with dimensional weight calculated as length by width by height in centimetres divided by 5000 (EU rate card effective 1 February 2026). - Advertising is a third cost on the same unit: a Sponsored Products bid is spent after the referral fee and, for FBA sellers, after the fulfilment fee. - Sponsored Products can target specific ASINs, categories and brands, so your own detail page is a surface other advertisers can buy, and defending a listing is a standing cost rather than a one-off. - A ChatGPT ad carries its own destination URL, so the same budget can end on your own shop, where no marketplace referral fee applies, or on the Amazon listing if that converts better for you. ## Full text If you sell on Amazon, the advertising question is not which platform is better. It is where the next euro goes, given that the unit it is chasing has already paid a referral fee, usually a fulfilment fee, and sits on a detail page that other advertisers are allowed to target. Sponsored Products still does something no other channel does, and this guide says so plainly. But it is worth being precise about what a marketplace ad euro buys, what a ChatGPT ad euro buys, and which of the two ends somewhere you control. For the two platforms compared as advertising systems, targeting, bidding, formats and reporting, read [ChatGPT Ads versus Amazon Ads](https://adsonomy.com/guides/chatgpt-ads-vs-amazon-ads) instead. **Key takeaways** - Cost the Amazon unit fully before you set an ad budget: referral fee on the gross price, fulfilment fee by size tier and charged weight, storage, returns handling. The ad budget sits on top of all of it. - Do not fund a ChatGPT test by cutting Sponsored Products. The two reach different moments, and the campaigns defending your own listings are the last thing to touch. - Choose the destination per campaign. Your own shop gives you the pixel, the customer record and no referral fee. The Amazon listing gives you a page that already converts and a checkout shoppers trust. - If you send ChatGPT traffic to Amazon, accept that you are buying clicks, not measured orders: OpenAI's pixel cannot go on an Amazon detail page. - Judge the test on your own shop's contribution after your own costs, not on ACOS. The two percentages describe different cost stacks. ### What does an Amazon unit already cost before the first ad euro? Three fee layers are committed before you bid on anything. The referral fee is charged on the gross, VAT-inclusive selling price for consumer orders: under the schedule effective 5 January 2026 for amazon.de, .fr, .it, .es, .nl and .com.be, that is 15% for most general-merchandise categories including Home & Kitchen, Sports, Toys & Games, Pet Supplies and Office, 13% for DIY & Tools and 7% for Computers, Consumer Electronics, Tyres and Full-Size Appliances, with a minimum of EUR 0.30 per item. If you use FBA, a per-unit fulfilment fee follows, set by size tier and charged weight under the EU rate card effective 1 February 2026. Storage, returns handling and category surcharges come after that. Only then does advertising start. Two mechanics inside that schedule are worth checking against your own catalogue rather than assuming. Some categories flip the whole rate at a price threshold: Grocery & Gourmet, Vitamins and Supplements and Pet Food are charged at 5% when the total price is at or below EUR 10 and 15% above it, and Baby, Beauty and Health & Personal Care flip at the same EUR 10 point. Others are genuinely marginal, such as Jewellery at 20% on the portion of the price up to EUR 250 and 5% on any portion above. A single blended percentage across a mixed catalogue misstates the fee at both ends of your price range, and Poland runs a separate, materially lower schedule. The fulfilment side has its own trap: dimensional weight is length by width by height in centimetres divided by 5000, and parcels and oversize are charged on the greater of actual and dimensional weight, so packaging moves the fee on its own. Storage is seasonal too, at EUR 27.54 per cubic metre per month from January to September and EUR 52.20 from October to December for standard-size non-apparel. All of these were checked in May 2026 and should be re-read in Seller Central before you model on them. ### Why does the next euro on Sponsored Products compete with your own listing? Because Sponsored Products can target specific ASINs, categories and brands, not only keywords. Your own detail page is therefore a targetable surface for any advertiser eligible in your category, and some of the placements a shopper sees on your page belong to somebody else. Spending on your own branded and defensive terms is not an optional growth activity in that situation. It is maintenance, and it has no natural end point. That is how a retail media auction is meant to work, and it is the reason the inventory exists at all. The practical consequence is that a rising share of the Amazon budget goes to holding ground rather than finding new demand, spent on a unit that has already paid a referral fee and a fulfilment fee. The marginal euro is doing less incremental work than a headline ACOS suggests, because part of it is buying a customer who was already looking at your product. The question is not whether to stop doing that. It is whether every additional euro belongs in the same auction, or whether some of it belongs earlier, before the shopper has decided where to buy. ### What does a ChatGPT ad euro buy that a Sponsored Products bid does not? It buys placement in a conversation where the shopper is still deciding what to buy, and it buys a destination you choose. A ChatGPT ad is a chat card or a product ad with its own target URL, so the same budget can end on your own shop rather than on a marketplace listing. Targeting is not a keyword list: an ad group carries up to 2,000 free-text context hints matched by meaning, which describe the situation someone is in rather than the words they typed. For a marketplace seller, the hint model is the genuinely new part. On Amazon you bid against a search term that already contains a product decision. A hint describes the problem before the product has been chosen: a coffee grinder that will not clog with oily beans, a child's bike for a first ride without stabilisers, a way to stop a bathroom fan dripping. There is no prompt-level report, so the hints are your only relevance lever and they have to come from real material, your own product pages, reviews and support questions; how to write them is in [contextual targeting and context hints in ChatGPT ads](https://adsonomy.com/guides/contextual-targeting-and-context-hints-in-chatgpt-ads). Two limits belong in the same breath: custom audiences are not available for campaigns targeting the EEA or Switzerland, so a European seller cannot rebuild an Amazon-style remarketing audience here, and conversion pixels, Conversions API keys and conversion-optimised bidding are capabilities gated per ad account and unlocked by OpenAI. ### Should the ad point at my own shop or at my Amazon listing? Decide it per campaign, not once for the account. Your own shop is the destination where no marketplace referral fee applies, where you can install OpenAI's conversion pixel or send server-side events through the Conversions API, and where the customer record is yours. The Amazon listing is the destination with a page that already converts, reviews that already exist and a checkout the shopper already trusts. Neither is automatically right, and for most sellers the shop destination is worth testing precisely because it has never been tested. The trade is concrete. Send the click to your own shop and you keep the measurement loop: order_created as a conversion event, order_created_attributed_sales and order_created_roas in insights at ad account, campaign, ad group and ad level, with the click window set on the conversion event itself. Send it to the Amazon listing and that loop is gone, because you cannot place OpenAI's pixel on a detail page you do not control. You are then buying clicks and reading whatever Amazon itself reports for off-Amazon traffic, which is a weaker measurement. Two notes before committing to the marketplace destination. The target URL has to be reachable by OpenAI's crawler and every creative goes through ad review, so whether a marketplace URL is acceptable is a policy question to settle against OpenAI's current ad policies rather than an assumption to build on. And a product ad is built from a feed item in OpenAI's own Agentic Commerce schema, a separate object from your Amazon catalogue, so that data has to be transformed before it uploads, as set out in [OpenAI product feeds versus Google Merchant Center feeds](https://adsonomy.com/guides/openai-product-feed-vs-google-merchant-center). Comparison: the same unit, two destinations for a ChatGPT ad. - Marketplace referral fee: charged as on any Amazon order, 15% of the gross price in most categories, 7% in Computers and Consumer Electronics, minimum EUR 0.30 per item under the schedule effective 5 January 2026 (Amazon listing) versus none, with your own payment processing, fulfilment and return costs instead (your own shop). - Fulfilment: FBA per-unit fee by size tier and charged weight, or your own FBM cost (Amazon listing) versus your own or your logistics provider's cost (your own shop). - Conversion measurement: clicks only from OpenAI's side, the pixel cannot be installed on a detail page (Amazon listing) versus pixel or Conversions API, with order_created and order_created_roas in insights (your own shop). - Bid optimisation: clicks or impressions (Amazon listing) versus conversions, once that capability is enabled on the ad account (your own shop). - Customer record: Amazon's (Amazon listing) versus yours (your own shop). - Landing page work: already done, and shared with competing placements (Amazon listing) versus yours to fix, and yours alone (your own shop). ### How do I measure a channel that can end in a marketplace I do not control? Split the measurement by destination and stop trying to produce one number. Own-shop campaigns get measured like any other paid channel into your shop: conversion events, a stated click window and a check against total orders. Campaigns ending on Amazon are measured in clicks out plus whatever Amazon reports for off-Amazon traffic, and any total-business read is directional rather than attributed. Do not convert ChatGPT results into ACOS to compare them with Amazon campaigns. An Amazon ACOS is calculated on revenue that has already paid a referral fee and a fulfilment fee, while a sale on your own shop carries a different cost stack entirely, so compare contribution after your own costs instead. The tracking template makes that easier: a query string template can be set at account, campaign, ad group or ad level, with macros for campaign, ad group, ad and ad account plus an opportunity reference, which puts clean parameters into your own analytics without hand-building URLs per ad. Setup is in [conversion tracking for ChatGPT ads](https://adsonomy.com/guides/openai-ads-conversion-tracking-pixel-and-conversions-api), and holdouts and geo splits are in [measuring ChatGPT ads with GA4, UTMs and incrementality](https://adsonomy.com/guides/measuring-chatgpt-ads-ga4-utm-incrementality). ### What is Sponsored Products still better at, and what should I not move? Sponsored Products captures demand at the moment of purchase, inside a marketplace the shopper opened in order to buy, on a page where the reviews, the delivery promise and the checkout are already in place. Nothing in this guide replaces that, and no ChatGPT campaign should be funded by switching it off. The campaigns holding your branded terms and defending your own detail pages are the last budget line to touch, not the first. Be specific about what you would be giving up. Sponsored Products bids against a search term that already encodes a product decision, and it has both a search term report and negative keywords, so you can see exactly what was searched and cut waste precisely. ChatGPT ads has none of that: no prompt-level data, no search term report and no documented negative list. If your current problem is that a proven Sponsored Products campaign runs out of budget before the end of the day, the answer is to raise that budget, not to start a new channel. The case for adding ChatGPT ads is not that Sponsored Products is bad. It is that Sponsored Products can only reach people who are already on Amazon, and part of your catalogue's demand is decided before anybody opens a marketplace at all. ### How should an Amazon seller run a first test without cutting anything? Fund it as an addition from a separate test budget, run one destination per campaign so the measurement stays clean, and set the decision criteria before launch rather than after. Start with your own shop as the destination if you have a shop worth sending traffic to, because that is the version you can actually measure, and because it is the question you have never been able to ask on Amazon. Pick a small set of products where the buying decision genuinely involves advice: something with a fit, sizing, compatibility or use-case question attached, not a repeat-purchase commodity. Write context hints from your own reviews and support questions. Set one conversion event, order_created, and one click window, and leave both alone for the length of the test. Size the daily number so it can run unattended: a ChatGPT campaign budget is an average, and OpenAI's Ads Manager states that maximum daily spend is twice the daily budget and maximum seven-day spend is seven times it, so a EUR 65 daily budget can reach EUR 130 on a single day. A full test design is in [how to plan a ChatGPT ads test budget](https://adsonomy.com/guides/testing-chatgpt-ads-budget-plan), and if your shop runs on Shopify, WooCommerce or BigCommerce the feed and pixel setup is in [ChatGPT ads for Shopify, WooCommerce and BigCommerce](https://adsonomy.com/guides/openai-ads-for-shopify-woocommerce-bigcommerce). Adsonomy runs the ChatGPT side only. It builds product ads from your feed and chat cards for the advice-stage conversation, and in Autopilot, Nomy proposes and applies changes within the guardrails and policy you set. It does not manage Amazon Ads campaigns, it does not touch your Seller Central account, and it is not affiliated with Amazon or OpenAI. ## Frequently asked questions Q: Does advertising in ChatGPT avoid the Amazon referral fee? A: Only for orders that complete somewhere other than Amazon. A ChatGPT ad that sends a shopper to an Amazon listing produces an Amazon order, which pays the referral fee like any other: 15% of the gross price in most categories and 7% in Computers and Consumer Electronics, with a minimum of EUR 0.30 per item, under the schedule effective 5 January 2026 for amazon.de, .fr, .it, .es, .nl and .com.be, checked May 2026. Q: Should I cut Sponsored Products to fund a ChatGPT test? A: No. Sponsored Products reaches a shopper who has already opened Amazon with a purchase in mind, and nothing in ChatGPT ads replaces that moment. Fund a ChatGPT test from a separate test budget and leave the campaigns that defend your own listings alone. Q: Can I point a ChatGPT ad straight at my Amazon listing? A: Each ad carries its own destination URL, which OpenAI's crawler has to be able to reach, and every creative goes through ad review before it serves. Whether a given marketplace URL is accepted is a review and policy question, so read OpenAI's current ad policies before building a campaign around that destination. You also cannot install OpenAI's conversion pixel on an Amazon detail page, so that campaign is measured in clicks. Q: Is conversion-optimised bidding available on a new ChatGPT ads account? A: Not automatically. Conversion pixels, Conversions API keys and conversion-optimised bidding are capabilities gated per ad account and unlocked by OpenAI, so a new account may have to start on clicks or impressions and add conversion optimisation once those capabilities are enabled. Q: How much should I budget for a first ChatGPT test as an Amazon seller? A: Treat it as an addition rather than a transfer, and size the daily number so it can run unattended. A ChatGPT campaign budget is an average daily amount: OpenAI's Ads Manager states that maximum daily spend is twice the daily budget and maximum seven-day spend is seven times it, so a EUR 65 daily budget can reach EUR 130 on a single day.