Advertising in ChatGPT: how OpenAI Ads works, and who it pays off for

OpenAI Ads went live for advertisers in the Netherlands on 1 September 2026, in beta, and is available in many other markets too. For the first time, you can advertise in ChatGPT yourself. Ads have appeared for Dutch users since 24 August 2026.

Advertising in ChatGPT does not start with your budget, but with two questions: does the channel fit your brand, and can you measure what it delivers? This article explains how OpenAI Ads works, who it already has value for, and why measurement differs per sales channel.

It is a beta. Features, countries and terms change quickly. What follows comes from the official OpenAI documentation. Where it helps, we link it to your ecommerce strategy.

What is advertising in ChatGPT?

An ad can appear below a ChatGPT answer. It is visible to users on the free version and on ChatGPT Go. On the paid plans, such as Plus, Pro and Business, no ads appear.

An ad consists of a brand name, a logo, a title, a short piece of copy, an image and a link to your destination. It is labelled as sponsored and sits apart from the answer. ChatGPT's answers stay independent of advertisers.

You pay per click or per thousand impressions. The actual cost depends on things like competition and relevance. There is no fixed price.

No keywords, but context

This is the most important difference from Google, Amazon or bol. OpenAI Ads does not buy keywords. The platform links an ad to the context and intent of an ongoing conversation.

You give direction with context hints: descriptions of situations where your product can be useful. That is steering, not an exact match, and no guarantee of being shown. Someone who asks for a recipe may see a relevant ad. You do not decide the exact moment.

There is also no keyword-volume planner and no search-term report like the ones you know from other channels. You work with your own demand data, with what you know from your webshop, reviews and customer questions, and you test that with real campaigns.

The practical consequence: you build around questions, situations and decision moments. You do not copy a keyword structure from another channel one to one.

Which ad formats are there?

There are two ways to create ads.

Standalone ads. You create a title, a short piece of copy, an image and a link yourself, as in the example above. Handy when you want to highlight a few products or offers.

Product feeds. You upload your product catalogue as a feed, and OpenAI builds ads from it at scale, with a product image, title and price. Useful with a large or frequently changing range. You deliver the feed in a Google-compatible product-feed structure, through an upload, a fixed URL or SFTP. Products expire after two weeks without a refresh, so for a continuous campaign you automate the delivery. OpenAI itself calls feed ads among its strongest performers so far.

How to set up a campaign

The structure has three levels: campaign, ad group and ad.

Campaign. Here you set your goal and budget: your objective, a budget (a total amount or a daily budget), a runtime and the countries where your ads may appear. Optionally you also choose the platform (iOS, Android or web). Keep one campaign to one business goal.

Ad group. One theme or need, with context hints that describe when your product is useful. Create several ad groups to steer different themes separately.

Ad. The title, copy, image and link the user sees. Create several per ad group, so you can compare different messages.

Note: your objective and your budget type are fixed once created. Choose wrong, and you make a new campaign. So decide them deliberately beforehand.

Writing context hints

Context hints are the core of steering on intent. You describe when your product is useful: the situation, the need, the moment. Not loose keywords.

Ask yourself: what else should someone know about this product, who does it help, and when is it useful? Compare:

• Weak: "running shoes"
• Strong: "cushioned everyday running shoes for beginners training for their first 5K"

Good context hints add something beyond your title and copy, connect a product feature to a real need, and describe situations rather than only broad audience labels. Use natural phrases, and only genuine, relevant cases.

Important: context hints are not an exact match and no guarantee of being shown. Someone does not have to use your exact words; the system recognises broader needs and variations.

Bidding and paying: CPM, CPC and oCPC

Your objective determines how you pay and what the system optimises for. There are three options.

• CPM. You pay per thousand impressions. For reach and visibility.
• CPC. You pay per valid click. For traffic to your site or product.
• oCPC. You still pay per click, but the system optimises toward a conversion you choose. This only works if you can measure that conversion, so with a pixel on your own site.

For your budget you choose a total amount or a daily budget. A daily budget is an average over seven days, so daily spend can fluctuate somewhat. There is a minimum daily budget that varies by currency; you see the current amount in Ads Manager. Feel free to start with a daily budget, so you can follow spend and results before you scale up further.

Where do you send traffic: your own site or a marketplace?

OpenAI Ads works best when you can connect a click to what it delivers. But the destination is also an ecommerce choice, not just a measurement question. Each route has a different balance between margin, data and conversion.

Your own webshop. The sale, the customer and the data are yours. Your gross margin is higher, because you pay no marketplace commission. You measure fully, with a pixel and server-side events, and this is the only route where OpenAI Ads' own conversion optimisation works. The downside: the conversion rate on your own site is usually lower than on a marketplace. Benchmarks for own webshops sit around 1% to 2.5%, Amazon around 10% to 15%. Whether that difference also holds for visitors who come in via AI is still unknown. That is new traffic without a benchmark.

Amazon. The conversion is usually higher, thanks to existing trust and purchase intent. Your gross margin is lower due to commission and fulfilment costs, and the data is not yours. Even so, Amazon is attractive for two reasons. With Amazon Attribution you can attribute external traffic to detail-page views, add-to-carts and purchases, on Amazon's side, not in OpenAI Ads' own optimisation. And sales from external traffic raise your sales velocity. That mainly feeds through into your category rank (BSR); the effect on your organic position per keyword is more indirect, because external traffic is not tied to a specific search term. For brands in Brand Registry, that traffic can also earn a Brand Referral Bonus, a discount on your selling fees. Note: this only applies to traffic that actually converts. Traffic that clicks away immediately adds no velocity.

Bol. The conversion is usually higher than on your own site too, but here a comparable attribution tool is missing. You see the clicks, not the revenue. For bol, OpenAI Ads is therefore more of a brand and demand channel than a direct-response channel.

So the choice is not a measurement question alone. It is an ecommerce trade-off: where does an extra visitor deliver most, given your margin, your measurement and your ranking goals? Often the answer is a combination, not either-or.

When not to start yet

Not every brand needs to start now. There are good reasons to wait:

• The margin is unknown. Then you do not know what a click or order may cost at most.
• The only destination is a channel where you cannot measure the revenue, while you do want to steer on it.
• The product or claim is not allowed or cannot be substantiated.
• There is no budget to test several messages. One generic ad then has to prove everything, and that rarely works.
• The landing page is not in order: wrong price, stock, or a page that does not deliver on the promise.

How to begin

A practical starting path, in the order that pays off most:

1. Work out your margin. What may a click or an order cost, given your margin? That limit decides whether advertising can contribute, before the first ad is even live.
2. Choose your destination. Your own site for full measurement and margin, Amazon for reach with measurement via Attribution, or bol as a brand and demand channel. Often it is a combination.
3. Set up your measurement. A pixel and server-side events on your own site, or an Amazon Attribution tag. Without measurement you steer blind.
4. Write context hints per situation. What is your product useful for, and when? Think in needs and decision moments, not in keywords.
5. Create several ads per theme and test. Change one thing at a time and read the outcomes as direction, not as a hard result. It is not a clean A/B test: the system splits delivery by expected relevance.
6. Start small and scale what works. Begin with a daily budget, follow the numbers and expand where the economics stay healthy.

Because OpenAI publishes no general benchmarks, you decide for yourself along the way what good numbers are, per country, category and type of question. In the end it is about visibility, conversion and margin, just like your other channels.

The EU context and consent

For brands in the Netherlands there is a concrete point of attention. In the EEA, personalised ads and matching your own customer data are limited at first. Targeting here leans more heavily on the context of the conversation.

On top of that, the measurement pixel requires valid consent up front. So you arrange consent and privacy before you start measuring, not afterwards. Have the legal side reviewed by whoever is responsible for it.

Frequently asked questions

What is advertising in ChatGPT? It is paid visibility within ChatGPT through OpenAI Ads. An ad can appear below an answer for users on the free version and on ChatGPT Go. It is labelled as sponsored and does not influence the answer itself.

Who sees the ads? Users on the free version and on ChatGPT Go. On paid plans such as Plus, Pro and Business, no ads appear.

Does OpenAI Ads work with keywords? No. The platform links ads to the context and intent of a conversation. You give context hints as direction, but those are not exact keywords and no guarantee of being shown.

Can you send traffic from ChatGPT to Amazon or bol? Yes. OpenAI Ads is its own channel, but you can have an ad link to an Amazon or bol listing. The difference is in measurement: external traffic to Amazon you measure with Amazon Attribution, while bol has no comparable tool. Your own webshop always gives the full picture.

Can you A/B test in OpenAI Ads? Partly. You can put several ads and context hints side by side, but it is not a clean split test. The system steers delivery by expected relevance, so variants do not get an equal, random split. There is also no separate experiment feature. So label each variant, change one thing at a time and read the outcome as direction. Testing structural choices, such as webshop versus marketplace, usually pays off more than micro-variants of a single ad.

What does advertising in ChatGPT cost? You pay per click or per thousand impressions. There is no fixed price. The cost depends on things like competition and how relevant your ad is to the conversation.

Is OpenAI Ads suitable for my brand? That depends on your ecommerce strategy. With your own webshop you measure most fully and keep more margin; a marketplace often converts higher and external traffic can help your category rank. Often a combination of destinations is strongest.

Is this the same as advertising on Amazon or bol? The basics look similar, but the logic differs. Marketplace advertising works with keywords and product demand. OpenAI Ads works with conversation context and intent. You do not carry a structure from one channel over to the other.

In closing

Advertising in ChatGPT is real and available now. The gain is in the choice up front: does this channel fit your brand, and can you measure what it delivers?

For brands with a strong webshop it is a serious new channel to learn from early. For marketplace brands it depends on the destination: to Amazon it is measurable via Amazon Attribution, to bol you deliberately accept weaker measurement.

Want to weigh this up for your own situation, and see how it fits your ecommerce strategy? Book a call. In our client cases you can see how we steer on visibility, conversion and margin.

Sources

• Ads in ChatGPT: The Basics · OpenAI Help Center
• Ads in ChatGPT · OpenAI
• Ads Manager Availability · OpenAI Help Center
• Create Campaigns for ChatGPT Ads · OpenAI Help Center
• Create Ad Groups for ChatGPT Ads · OpenAI Help Center
• Create Campaigns from Product Feeds · OpenAI Help Center