Michael King keeps saying ChatGPT ads don't work, and the new Schulze/Kaiser pap
By Matt Hammel · November 6, 2026 · Curated by George's Blog
Michael King keeps saying ChatGPT ads don't work, and the new Schulze/Kaiser paper offers some real ammo: a paid ChatGPT visit drove $0.89 in revenue vs. $2.45 for a paid Google search visit.
https://lnkd.in/ggYAm4ix
I believe the data. I just don't think it tells you where this ends up.
Every major ad platform looked rough in its early years.
Google AdWords launched in October 2000 selling impressions off a CPM rate card. The pay-per-click auction that ranked ads by bid and click-through rate didn't show up until 2002. Quality Score came in 2005. Revenue went from $19M in 2000 to $1.47B in 2003.
Facebook's 2012 IPO filing said it did "not currently directly generate any meaningful revenue" from mobile. Mobile was 23% of its ad revenue in Q4 2012 and 53% a year later.
In 2019, buyers described TikTok's self-serve ads as buggy, with few metrics. By 2021 it was doing roughly $4B in ad revenue.
The audience shows up first, then pricing, targeting, measurement and tooling get rebuilt in public. OpenAI, despite some real turnover, has hired the best of the best from the previous ways of new dominant ad platforms. Not to mention they also have machine-god intelligence to help them build it.
ChatGPT ads went live on February 9. In under nine months they've moved from $60+ CPMs to CPC and outcome-based bidding, added a pixel, Conversions API and custom audiences, and passed a $1B run rate. And the inventory is people describing exactly what they need, in their own words. Once the system learns to price that intent, it will be formidable.
Brands today should be running fast-cycle, structured tests while auctions are thin and learning is cheap.
Writing off ChatGPT ads at month nine is like writing off AdWords in 2001. Who's testing, and what are you seeing? What's working? What's not?