Imagine a shopper typing, “I need a waterproof backpack under $150 for a three-day trip,” and your product showing up right inside the answer. That’s no longer a thought experiment. This is the first holiday season when anyone can buy ads in ChatGPT, so let’s talk about whether you should try it and how to do it without wasting money.
What’s Actually New
OpenAI confirmed in January that it was testing ads for US users on the Free and Go tiers. They went live in February, and in May OpenAI opened a self-serve Ads Manager in beta, added cost-per-click bidding, and rolled out pixel and Conversions API measurement. Ads appear as labeled units below ChatGPT’s answers, and Plus, Pro, and business accounts don’t see them.
Why does the timing matter? Adobe expects AI-driven traffic to US retail sites to rise 130% this holiday season, with the biggest jumps on Thanksgiving, Black Friday, and Cyber Monday. More shoppers are starting their search in a chat window, and that’s where these ads live.
Who Should Test It (And Who Can Wait)
Honestly, it won’t suit everyone. Here’s a simple way to decide.
A good fit if you:
- Sell products people research before buying, like electronics, home goods, or gear
- Have clean product data and good images
- Have a stable, profitable Amazon or website business and a small budget you can afford to experiment with
Better to wait if you:
- Are still sorting out inventory or shipping for the holiday rush
- Can’t yet measure sales from outside channels
- Would need to pull budget from campaigns that are already working
Your Product Data Is the Real Entry Ticket
The ad format most relevant to sellers is the product feed ad, which can show images, titles, star ratings, prices, sale prices, and brand. OpenAI says feed-based ads have been among its best performers so far, though there’s no independent data to back that up yet. Treat that as a claim to test, not a promise.
Either way, your feed has to be right. Feed items expire after two weeks, so a stale price or out-of-stock product can disappear or mislead shoppers. Solid Ecommerce Product Information Management means one clean source of truth for titles, attributes, and prices, so the same details show up everywhere your product does. If your Amazon listing says one thing and your site says another, an AI system has to guess, and guessing is how products get left out of answers.
How to Run a Small, Smart Test
You don’t need a big budget. One guide lists a minimum of $25 a day per campaign in the US, which makes a short test affordable. Here’s a simple plan:
- Pick three to five products with healthy stock and good margins. Skip anything you might run out of before December.
- Set a fixed test budget and an end date. Two to three weeks is plenty to learn something.
- Add measurement first. Set up the pixel or Conversions API before you spend, so every click can be traced to a sale.
- Compare against your baseline. Check cost per sale and profit after fees, not just clicks.
- Write down what you learn whatever the result, because next year’s plan will be much easier.
Keep in mind that OpenAI gives advertisers aggregated results, not individual conversations, so you’re judging performance on outcomes rather than on what shoppers said.
Don’t Forget the Free Side of AI Shopping
Ads are only one way to show up. ChatGPT’s shopping research feature, which builds buyer’s guides, draws from organic results rather than paid placement. That means AI Marketplace Search rewards the basics: complete listings, clear specs, honest reviews, and consistent information. Fixing those helps you in the ad auction and in unpaid answers at the same time.
Treat New Channels Like Experiments
The sellers who benefit most from new channels tend to treat them as experiments inside a bigger plan. A Global Ecommerce Accelerator mindset means every new channel earns a small test budget, a clear success number, and a stop date. If it works, you scale it. If it doesn’t, you’ve paid a small price for a clear answer.
That same Global Ecommerce Accelerator discipline applies to your data. Keep one master version of your product information, check it weekly during peak season, and make sure what an AI sees matches what a shopper sees. When the next channel arrives, and one always does, you’ll be ready to test it in days instead of weeks. Strong AI Marketplace Search visibility and clean Ecommerce Product Information Management are what let a Global Ecommerce Accelerator approach move fast.
Sources
- Ads in ChatGPT: self-serve Ads Manager (beta), cost-per-click bidding, pixel and Conversions API measurement, and aggregated results rather than individual conversations (May 5, 2026): OpenAI, “New ways to buy ChatGPT ads”
- Self-serve and CPC bidding coverage (May 13, 2026): ContentGrip
- January test announcement for US Free and Go users: Adventure PPC
- February go-live and May self-serve opening: Stub Group
- Ads shown only to Free and Go users, product feed ad details, feed items expiring after two weeks, OpenAI’s own claim about feed ads, and the $25 per day US minimum: Everything-PR guide
- AI traffic to US retail sites up 130% this holiday season: Digital Commerce 360, citing Adobe
- Shopping research drawing on organic results, not paid exposure: Design Compass