OpenAI announced on September 1 that ChatGPT Ads had crossed $1 billion in annualized revenue. The platform launched less than 200 days ago.
That number matters less as a milestone than as a signal. When a new ad platform reaches $1 billion in run rate, the volume of advertisers competing for inventory starts rising fast. The low-CPM window — the period where you're paying what early adopters pay, not what a crowded auction charges — closes once mid-market and SMB demand arrives at scale.
SMBs are already there. OpenAI confirmed they represent a "material share" of the advertising business, driven by self-serve access through Ads Manager.
What changed on August 31
Until August 31, self-serve ChatGPT Ads Manager was limited to US advertisers. That changed: India, 31 European markets, and MENA all got self-serve access simultaneously, bringing the total to 40+ countries. Managed service had already operated in some of these markets through agencies, but direct self-serve is different. It removes the minimum spend floor, removes the agency requirement, and opens the platform to local brands who were previously priced out.
For established advertisers, this matters competitively. More self-serve demand in 40 countries means the auction fills faster, inventory pressure builds, and the efficiency that characterized early test results becomes harder to replicate.
The MENA expansion is worth flagging separately. According to Communicate Online, OpenAI highlighted this market specifically — suggesting there's both platform infrastructure and advertiser interest there. If any portion of your addressable market sits in the Gulf or broader MENA region, check that your conversion tracking covers these markets before running spend there.
The platform can now run a real direct-response campaign
Early ChatGPT Ads coverage — including the May 2026 period when self-serve first opened — treated the platform as a branding experiment. The arguments for caution were reasonable: no conversion optimization, no custom audiences, attribution was thin. Those arguments have eroded significantly.
CPA bidding is live. Custom audiences from hashed email and phone lists are available. Geo-targeting works down to state and DMA level in the US. The Conversions API and OpenAI Pixel are both functional. Third-party measurement integrations exist. According to Search Engine Land, OpenAI also shipped a batch of additional ad stack updates alongside the $1B announcement — spanning audience management, campaign pacing, and reporting.
This is no longer a place you go to run brand impressions at a premium. You can run a direct-response campaign with a feedback loop that resembles, at least structurally, what you'd expect on Meta or Google.
What's still missing: the depth of audience segmentation, years of purchase-intent signal, and the creative format variety that make Meta and Google efficient at scale. Those gaps are real. They justify running smaller test budgets and controlled experiments — not skipping the platform entirely.
The CPM math heading into Q4
ChatGPT Ads launched into a relatively low-competition environment. The $1B run rate, the 40-country expansion, and the SMB influx all point in the same direction: more demand, same inventory, upward pressure on clearing prices.
This pattern has played out on every ad platform that expanded self-serve access internationally. It's not guaranteed, and it's not linear. But the accounts testing in September are doing so at lower average CPMs than the accounts testing in November, when Q4 auction pressure compounds with the broader expansion effect.
There's also an asymmetric benefit to running in September specifically. ChatGPT Ads' CPA bidding model learns from your conversion data — the longer it runs, the better it performs. Starting in September gives your campaigns a six-to-eight week learning head start before Black Friday CPMs peak across every platform simultaneously.
What a useful test looks like
If you haven't run ChatGPT Ads yet, the setup that makes sense right now is simple: one campaign, CPA bidding objective, your highest-margin product or category, US targeting where the audience signals are deepest, and a 30-day budget you can treat as a learning cost.
Wire up conversion tracking before you launch. The attribution gaps that made early tests look inconclusive were largely a setup problem, not a platform problem. A live pixel or Conversions API integration is not optional — without it, the CPA bidding model has nothing to optimize toward and you get CPM-equivalent spend without the optimization benefit.
Don't expect ChatGPT Ads to replace Meta or Google in Q4. The platform has hundreds of millions of monthly actives, but the audience skews toward information-seekers in commercial research phases. That's valuable for considered purchases, subscription products, and anything with a longer decision cycle. It's weaker for impulse categories.
If you're already running ChatGPT Ads and want to see how it stacks up against your other channels on a true cost-per-outcome basis, the free audit at Gromerce shows spend efficiency across your active platforms in a few minutes.
The $1 billion run rate is real. So is the crowd of advertisers walking in behind it.
Sources: OpenAI blog, Search Engine Land, September 2026
