# Marketplace ads vs ChatGPT ads: where to spend the next euro

URL: https://adsonomy.com/guides/marketplace-ads-vs-chatgpt-ads
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 with any marketplace.

## Summary

The hub page for the marketplace advertising question: on a marketplace you pay a commission to sell there and then pay again to be seen there, in an auction fought against sellers of the same item and sometimes against the marketplace's own brands. Explains what is different about reaching a shopper inside a ChatGPT conversation, where the question is still what to buy rather than which of several identical listings, and is explicit about where marketplace advertising wins, which is capturing demand at the moment of purchase, and about the fact that a marketplace owns the customer relationship either way. Fee numbers belong on the per-marketplace pages; this page is the reasoning.

## Key facts

- A marketplace charges a commission on the order whether or not an ad produced it, and then sells visibility on the same platform separately, so ad spend comes out of the margin the commission has already reduced.
- Marketplace ad auctions are usually fought over the same catalogue item, often the identical listing, where the visible differences are price, delivery date and rating, so the bid becomes the main lever.
- Many marketplaces also sell their own brands and their own retail stock, which means the party setting the auction rules can also bid in it.
- ChatGPT ads has no keyword list: an ad group carries up to 2,000 free-text context hints matched by meaning, and there are two creative types, a chat card and a product ad built from a feed item.
- A ChatGPT ad's destination is a target URL you supply, so the click lands on your own site with your own analytics and conversion tracking; a marketplace ad keeps the session, the basket and the order inside the marketplace.
- No marketplace commission is taken from a sale that completes on your own checkout, so the allowable cost per order is higher there: on an item selling at EUR 60 with a 15 percent commission it is EUR 18.00 against EUR 25.50, and that allowance rather than the cost per click decides which channel is cheaper.
- Custom audiences are not available for campaigns targeting the EEA or Switzerland, and the platform has no language targeting field at all.
- Marketplace advertising captures demand at the moment of purchase, where the account, the stored card, the delivery promise and the reviews are already in place, and a marketplace order stays the marketplace's customer relationship whichever channel created the demand.

## Full text

Every marketplace seller eventually runs the same sum. You already pay the marketplace a commission on each order. Then you pay the same marketplace again to be seen on it, in an auction where most of the other bidders are selling the same item you are. This guide is not a fee table: commission rates, fee bases and ad formats differ per marketplace and belong on the page for that marketplace. It is the reasoning that is the same everywhere, and it ends with a way to decide which side of the split the next euro belongs on.

**Key takeaways**

- Work out the contribution margin per item after commission, fulfilment and returns before comparing channels: that number, not revenue, is the ceiling on what any ad can cost you per sale.
- Separate defensive marketplace spend, on your own branded searches and your own product page, from spend that creates new demand. Only the second is comparable with a ChatGPT budget.
- On a shared listing your bid is close to the only lever you have, so an auction that keeps getting more expensive is a structural feature, not a badly built campaign.
- Compare allowances, not clicks. A sale that completes on your own checkout carries no marketplace commission, so it is allowed to cost more to win and still leave you more money, and that gap widens with every point of commission your category carries.
- Fund a ChatGPT test from its own line with its own floor and conversion tracking live before the first euro is spent, rather than by cutting a marketplace campaign that is still profitable.

### On a marketplace you pay once to sell there and again to be seen there

A marketplace takes a commission on every order, and it takes it whether the sale came from an ad, from organic position or from a shopper typing your brand name. Advertising is a second, separate payment to the same platform for visibility on it. The ad budget therefore comes out of what is left after commission, fulfilment and returns, which is why an ad cost of sale that looks comfortable against revenue can quietly be most of the contribution margin.

That is the sum worth doing before any channel comparison. Revenue is the wrong denominator: the number that matters is what one more unit actually contributes after every platform cost, because that is the ceiling on what a sale is allowed to cost to acquire. Check the base your marketplace uses as well, since some calculate commission on the full amount the shopper paid, VAT included, and others on a net figure. The same headline percentage is a different cost depending on which one applies to you.

The structural point follows from there. The platform that sets your cost of sale is also the platform selling you the remedy for weak visibility. When a commission changes or a new fee appears, the advertising line is usually the only one a seller can move inside a quarter, so it absorbs the change. That is not a conspiracy, it is the shape of the market you are trading in, and it is worth naming before you decide the next euro has nowhere else to go.

### Marketplace ad auctions are fought against sellers of the same item, sometimes including the marketplace's own brands

A marketplace ad usually appears beside, above or inside the same product sold by somebody else, and on a catalogue marketplace it can be the identical listing with a different seller attached. What a shopper can see between the offers is price, delivery date, seller rating and sometimes nothing at all, so the bid turns into the main lever. Many marketplaces also sell their own brands and their own retail stock, which means the party that sets the auction rules and holds the demand data can be a bidder in the same auction.

The consequence for pricing is mechanical. The number of sponsored slots on a results page is roughly fixed while the number of sellers competing for them is not, so the clearing price drifts toward whatever the most margin-tolerant bidder can pay. That may be a seller with a lower landed cost, one funding growth deliberately, or one clearing stock at zero contribution. If your cost base is ordinary, you are bidding against a ceiling that is higher than yours, and no amount of campaign hygiene changes that. The documented mechanics of one such auction, including Featured Offer eligibility, match types and dynamic bids, are set out in [ChatGPT Ads vs Amazon Ads](https://adsonomy.com/guides/chatgpt-ads-vs-amazon-ads).

A large share of marketplace ad budget is also defensive. It buys back your own branded searches and keeps competitors off your own product page. That spend is often correct, and cutting it can cost more than it saves, but recognise it for what it is: rent on a position you already earned. It is not comparable with a budget that has to create demand from nothing, and mixing the two into one blended figure is how sellers end up unable to say what their advertising actually does.

### Cost per click is the wrong comparison, allowable cost per order is the right one

The reason a marketplace euro and a ChatGPT euro cannot be compared at face value is that they are spent against different amounts of remaining margin. A marketplace takes its commission out of every order before you have paid for any advertising at all. An order that completes on your own checkout is never charged that commission, so the same item leaves more money behind to pay for the sale that produced it. That gap, not the cost per click, is what decides which channel is cheaper for you.

Run it as a break-even. Work out what one order contributes in each place after every cost except advertising. Those two numbers are what a sale is allowed to cost before it stops being worth having. Put your own figures in; the ones below are placeholders chosen to show the shape of the sum, not a benchmark.

- Price the shopper pays, excluding VAT: 60.00 on the marketplace, 60.00 on your own store
- Cost of goods: 24.00 on the marketplace, 24.00 on your own store
- Pick, pack, ship and expected returns: 9.00 on the marketplace, 9.00 on your own store
- Marketplace commission at 15 percent: 9.00 on the marketplace, none on your own store
- Payment processing: inside the commission on the marketplace, 1.50 on your own store
- Left to pay for the ad: EUR 18.00 on the marketplace, EUR 25.50 on your own store

On those numbers a marketplace sale stops being worth having once the advertising that produced it costs more than EUR 18.00, and the own-store sale at EUR 25.50. The ChatGPT sale is therefore allowed to cost about 40 percent more to win and leave you exactly as well off. Move the commission to 21 percent, which plenty of categories carry, and the marketplace allowance falls to EUR 14.40 while the own-store allowance does not move at all, so the same sale may then cost 77 percent more off the marketplace and still break even. The higher the commission on your category, the wider that gap, and it opens before anybody has bid anything.

Two further effects push the same way and one pushes back. In favour: the marketplace auction is priced by sellers of the identical item and it carries your defensive spend as well, while a ChatGPT ad group is matched to a described intent rather than to your SKU, so you are not bidding against a copy of your own listing. Also in favour: a customer who buys on your own checkout can buy again with no commission taken on the second order, which the first order alone never captures. Against: the volume available in a conversation surface today is nowhere near a marketplace results page, and no cost per click for it is published by anyone, ourselves included.

So the honest claim is not that ChatGPT ads is cheaper per click. It is that a sale won there is allowed to be considerably more expensive and still be the better buy, that the size of that allowance is arithmetic you can do this afternoon on your own margins, and that anything under it is cheaper in the only sense that pays: money left after the sale.

### Inside ChatGPT the question is still what to buy, not which of these identical listings

A marketplace results page is what somebody opens once they already know the category and want to compare offers within it. A conversation in ChatGPT often starts a step earlier, with a problem described in the shopper's own words and no shortlist yet. That changes what an ad can usefully say: the message that works is why this product suits what the person just described, not that it is a few cents cheaper than the identical listing beside it.

The documented mechanics follow that difference. ChatGPT ads has no keyword list and no match types. An ad group carries up to 2,000 free-text context hints matched by meaning, so the unit of competition is a description of an intent rather than a search term every seller of the same item is bidding on. There are two creative formats, a chat card with its own title, body and image, and a product ad built from an item in an uploaded feed. How the hints behave in practice is covered in [context hints in ChatGPT ads](https://adsonomy.com/guides/contextual-targeting-and-context-hints-in-chatgpt-ads), and what to write in the card itself in [writing chat card creative](https://adsonomy.com/guides/writing-chatgpt-ad-creative-chat-cards). The feed is a separate object from your marketplace catalogue and uses OpenAI's own schema, as explained in [OpenAI product feeds vs Google Merchant Center](https://adsonomy.com/guides/openai-product-feed-vs-google-merchant-center).

Be honest about what is not known. Nobody has a like-for-like cost per sale between a mature marketplace auction and this surface, and anyone quoting one is guessing. The available volume is also nowhere near what a marketplace results page delivers today. What can be settled in advance is the allowance, the sum set out above: how much a sale is allowed to cost in each place. Beyond that the claim is narrower and more useful than a price comparison, because the money buys a different thing, a mention while the choice is still open, in a place where your product is described rather than compared row by row against a copy of itself.

### A ChatGPT ad lands on a page you own, which is the part a marketplace never sells you

The destination of a ChatGPT ad is a target URL you supply, so the click arrives on your own site, inside your own analytics, with your own conversion pixel or server-side events and whatever consented email capture you run. A marketplace ad is built the other way around: the session, the basket, the order, the address and the follow-up all stay inside the marketplace, because keeping them there is the product the marketplace sells.

Two euros with the same immediate return are therefore not worth the same. One produces an order and nothing else. The other produces an order plus a visitor you can remarket to, a customer you can email with consent, and a data point that belongs to you. It is also the one you can measure properly: conversions are reported against the click window set on your conversion event, so choose that window deliberately and state it on every report. The practical setup is in [OpenAI Ads conversion tracking](https://adsonomy.com/guides/openai-ads-conversion-tracking-pixel-and-conversions-api), and the wider measurement picture in [measuring ChatGPT ads](https://adsonomy.com/guides/measuring-chatgpt-ads-ga4-utm-incrementality).

OpenAI has also described a Sponsored Agents format, in testing with selected US advertisers, where a click opens a clearly labelled conversation with a business's own agent. Treat it as a test to watch rather than something to plan a budget around.

### Marketplace advertising wins the moment of purchase, and that is worth paying for

The strongest argument for the marketplace euro is that everything between the click and the money is already solved there. The shopper has an account, a stored card, a delivery promise they trust, a returns policy they have used before and reviews they have already read. An ad on that surface captures demand at the point where a person is ready to pay, with none of the friction an unfamiliar shop has to overcome first.

There are second-order effects too. Marketplace search ranking generally responds to sales, so paid sales can lift the unpaid position of the same listing and keep lifting it after the campaign stops. A launch with no sales history often has no other way to start. And defending your own product page is something only an on-marketplace campaign can do. A ChatGPT ad has no equivalent effect on any marketplace's ranking.

So the answer to the pillar question is frequently the marketplace. If the item is a repeat purchase people type by name, if you are launching and need velocity, if a competitor is camped on your product page, or if the marginal ad cost of sale is still under your contribution margin, the next euro belongs where it already is. The argument in this guide is for a second line in the budget, not for abandoning the first.

### The marketplace owns the customer relationship whichever channel you paid for

A marketplace order is the marketplace's customer. That is true of a sale produced by an ad on the marketplace, and it is equally true of a sale your off-platform advertising produced if the shopper completed the purchase on the marketplace. Spending outside the platform does not move the relationship by itself. It only moves it when the shopper actually buys from you, on your own checkout.

That creates an attribution problem in both directions, and it is the honest weakness of any comparison including this one. Somebody who sees an ad inside ChatGPT may well go and search your brand on the marketplace afterwards and buy there out of habit. That sale appears in the marketplace's numbers and not in your ChatGPT reporting, so a naive read undercounts the conversation and overcredits the marketplace. Some marketplaces sell an off-site attribution product; check whether yours does and what it actually covers. Where it does not, holdouts and geo tests are the only clean read, which is why the measurement plan has to exist before the budget does.

It also sets a realistic goal. ChatGPT ads is not an escape route from a marketplace, and no sensible plan treats it as one. It is a way to reach demand that has not yet been shaped by a comparison page, and to capture part of it somewhere you keep the customer. If you have no own store that can take the order, most of that advantage disappears before you start; see [OpenAI Ads for Shopify, WooCommerce and BigCommerce](https://adsonomy.com/guides/openai-ads-for-shopify-woocommerce-bigcommerce) for what the own-shop side needs.

### What a European seller has to plan around before moving any budget

Two platform constraints matter immediately to a European marketplace seller. Custom audiences are not available for campaigns targeting the EEA or Switzerland, so a customer list, or a lookalike built from one, cannot be the foundation of a European plan. And the platform has no language targeting field at all, so a campaign cannot be split by the shopper's language the way a marketplace seller normally splits campaigns country by country.

What you do control is location and creative. Targeting takes location IDs from a published geo list, so country and region selection is precise, and language is handled by writing the ad and the landing page in one language and keeping that ad group to the market it is written for. That is a structure choice for control and readable reporting, not a targeting guarantee. Regional differences in how the platform operates are covered in [ChatGPT ads in the US, the EU and Asia](https://adsonomy.com/guides/chatgpt-ads-in-the-us-eu-and-asia), and the mechanics of the surface itself in [how ads in ChatGPT work](https://adsonomy.com/guides/how-ads-in-chatgpt-work).

### How to decide where the next euro goes

Decide per item and per margin rather than per channel. If the marginal ad cost of sale on the marketplace is still below the contribution margin left after commission, fulfilment and returns, that euro is profitable and should keep running. The case for a ChatGPT euro is strongest for products people ask advice about, for items where the marketplace auction has already reached or passed your margin ceiling, and for sellers with their own shop that can take the order and keep the customer.

- Calculate the contribution margin per item after commission, fulfilment and returns, using the commission base your marketplace actually applies. That number is the ceiling on your ad cost of sale.
- Check whether you are the only seller of the item or one of many on a shared listing. A shared listing means your bid is close to your only lever.
- Check whether the marketplace sells its own brand in your category, because that changes who you are bidding against.
- Split your marketplace spend into defensive and demand-creating. Compare only the second with anything off-platform.
- Ask whether your product is one people describe a problem about or one they restock by name. Advice-shaped demand is where a conversation ad has something to say at all.
- Fund the test from a separate line with its own floor and a fixed period, with conversion tracking live before the first euro is spent. A budget and test structure is set out in [how to test ChatGPT ads](https://adsonomy.com/guides/testing-chatgpt-ads-budget-plan).

Adsonomy runs the ChatGPT side of that split. Nomy, the AI manager, reviews performance, proposes changes and applies the ones you have allowed, within the guardrails and policy you set. It does not manage marketplace advertising, and Adsonomy is independent: it is not affiliated with OpenAI or with any marketplace.

Comparison: marketplace advertising versus ChatGPT ads.
- What the shopper is doing: comparing offers inside a category they have already chosen (marketplace advertising) versus often still working out what to buy, in their own words (ChatGPT ads).
- Who you outbid: sellers of the same item, often on the same listing, sometimes the marketplace's own brands (marketplace advertising) versus other advertisers matched to a described intent, not to your SKU (ChatGPT ads).
- What the ad can say: price, delivery and rating inside a fixed listing format (marketplace advertising) versus a chat card in your own words, or a product ad built from your feed (ChatGPT ads).
- Targeting unit: keywords, products or categories, depending on the marketplace (marketplace advertising) versus up to 2,000 free-text context hints matched by meaning (ChatGPT ads).
- Where the click lands: inside the marketplace (marketplace advertising) versus on a target URL you supply, on your own site (ChatGPT ads).
- Who owns the customer: the marketplace (marketplace advertising) versus you, if the purchase completes on your own checkout (ChatGPT ads).
- What you pay besides the ad: commission on the order, plus fulfilment and returns (marketplace advertising) versus your own fulfilment and payment costs, no marketplace commission (ChatGPT ads).
- Effect on organic position: paid sales generally feed marketplace ranking (marketplace advertising) versus no effect on any marketplace's ranking (ChatGPT ads).

## Frequently asked questions

Q: Should I move budget out of marketplace ads into ChatGPT ads?
A: Not as a swap. A marketplace campaign whose marginal ad cost of sale is still below the contribution margin left after commission, fulfilment and returns is profitable and should keep running. ChatGPT ads buys a different moment, earlier in the decision and on a page you own, so it is funded as a separate test line with its own floor and its own measurement rather than by cutting a campaign that still pays for itself.

Q: Why do marketplace advertising costs keep rising even when my listing gets better?
A: Because the number of sponsored slots on a results page is roughly fixed while the number of sellers bidding for them is not. In an auction fought over the same catalogue item, the clearing price moves toward whatever the most margin-tolerant bidder can pay, which may be a seller with a lower landed cost, a seller clearing stock, or the marketplace's own brand. A better listing improves your conversion rate, not the number of slots.

Q: Can I point a ChatGPT ad straight at my marketplace listing?
A: Technically the ad takes a target URL you supply, which has to be reachable by OpenAI's crawler, so a listing URL can be used. Practically you lose the measurement: you cannot install OpenAI's conversion pixel on a marketplace product page, and server-side conversions need order data the marketplace may not expose in a form you can match. You also hand the customer relationship back to the marketplace on a click you paid for yourself.

Q: Can I target by language or upload my customer list for a European campaign in ChatGPT ads?
A: No on both counts. Custom audiences are not available for campaigns targeting the EEA or Switzerland, and the platform has no language targeting field at all. You control the market through location targeting and you control language through the creative and the landing page, which is why European sellers usually run one ad group per market and per language.

Q: Is ChatGPT advertising cheaper per sale than marketplace advertising?
A: It can be, and you can work out whether it would be for you before spending anything. No marketplace commission is taken from a sale that completes on your own checkout, so that order is allowed to cost more to win and still leave more money behind. On an item selling at EUR 60 with a 15 percent commission, the marketplace allowance is EUR 18.00 against EUR 25.50 on your own store, so a ChatGPT sale may cost about 40 percent more and still break even. What nobody can tell you yet is the cost per click on the new surface, so run the sum on your own margins and then test against it.