Amazon and OpenAI announced ChatGPT Ads today. My feed is already full of takes.
Most of them skipped the two lines that matter.
First: it runs through Amazon DSP, as a managed service.
Closed pilot, US only. The first named tester is Delta Vacations, a travel brand.
So if you’re sitting down to plan your ChatGPT ad strategy this week, there’s nothing to plan yet. You can’t buy it. And in the pilot, Amazon’s team runs the campaigns, not you.
Worth noting where it landed. Every new surface Amazon has opened went into DSP first. Fire TV, Twitch, Prime Video, now this. Sponsored Products is the console most sellers live in. DSP is where the new inventory shows up.
Second, the question I haven’t seen anyone ask.
Amazon’s own announcement says nothing about measurement at all. The only reporting detail I could find is a spokesperson quoted in the trade press: “aggregated performance data including impressions, clicks, cost per result, CPM and CPC.”
Aggregated is not how DSP reporting usually reads.
If OpenAI is serving the ads, what does conversion reporting look like on your side? Nobody has said either way.
That’s what I’d want answered before moving budget. Not which product to test first.
Now the part I’d actually prepare for.
This traffic is going to convert well. Someone typing “cordless vacuum” into Amazon gave you two words. Someone in ChatGPT says “I have two cats, stairs, and almost no storage, what should I look for.”
They arrive having already told you what they need.
Then they land on a listing built to rank for two words.
The main image answers a keyword. The bullets answer a keyword. Nothing on the page confirms the three things they just said out loud.
That’s where it will leak. Not because the traffic is cold. Because the listing was built for a different shopper.
One of the two ad formats is a product feed ad that builds creative straight from your catalog. So your listing data is the ad whether you touch it or not.
The one place I’d expect it to underperform is research-mode conversations. “What should I look for in X” is someone building a shortlist, not buying today. Good intent, wrong week. Those clicks cost the same.
We’ve been buying off-marketplace traffic into marketplaces for a while.
One brand’s Walmart account in March: $7,857 in search engine ad spend, $68,830 back. 53% of that month’s sales came from people who never started on the marketplace.
The traffic was never the hard part. Matching the page to what made them click was.
Anyone been offered access to the pilot yet?
