OpenAI is now an advertising business. That would have sounded odd a year ago, when Sam Altman was openly skeptical about putting ads in ChatGPT. But the company is reportedly chasing $2.5 billion in ad revenue this year and far more by 2030, and in May it made its biggest move yet: a self-serve Ads Manager Beta that lets US businesses buy ads inside ChatGPT directly, with no agency and no enterprise-sized minimum to clear.
So we did what any performance team that lives and breathes this stuff would do. We got our US branch in early, spun up real campaigns with real budget, and spent weeks poking at every corner of the platform. Here’s what we found, who it’s actually built for, and where we think it fits in a modern media mix.
What the Ads Manager beta actually is
Ads Manager is OpenAI’s self-serve portal for creating, launching, and measuring ads that appear inside ChatGPT conversations. Think of it as the control room: you set objectives, upload creative, manage budgets, bid, and read your performance reports all in one place, without needing an agency middleman or a managed-service handshake with OpenAI.
The ads themselves show up as clearly labeled blocks beneath a ChatGPT response, with a subtle background so they never get confused with the actual answer. Instead of keyword targeting, the system matches your ad to the context and intent of the live conversation. That’s the part that makes this genuinely different from search or social. You’re not bidding on a query someone typed. You’re showing up in the middle of a decision someone is actively working through.
One important detail: ads only serve to Free and Go tier users, not Plus, Pro, or Business plans, and not to anyone predicted to be under 18. Delivery is currently limited to a handful of countries on the user side, so this is still an early, phased rollout. Treat it like the new channel it is, not a mature one.
Who it caters for
The removal of the old enterprise minimum reframes who this is for. It’s now open to US-based businesses of essentially any size, which is a big deal for small and mid-sized advertisers who never could have afforded the original pilot.
From what we saw, the sweet spot right now is high-consideration, research-heavy categories: B2B SaaS, professional services, education, and considered ecommerce. Those are the conversations where someone is comparing options and forming a decision in real time, which is exactly the moment a relevant ad earns its click. If you’re a purely local service business, or you operate in a heavily regulated category like healthcare or finance, you can still qualify in some cases, but you’ll want to move more carefully and read the eligibility rules closely. Political advertising isn’t supported at all.
If you’ve used Meta Ads Manager, you already know your way around
This was the biggest surprise, and honestly the most reassuring one. Once you’re inside, the platform feels a lot like Meta Ads Manager.
The structure is the familiar campaign, ad group, ad hierarchy. You pick an objective (Reach if you want impressions, Clicks if you want, well, clicks), set budgets and pacing, and bid at the ad group level. There’s geo targeting for regional and local campaigns, the kind of location control anyone coming from Google or Meta expects. Retailers can upload a full product feed and auto-generate ads from individual catalog items, which is the same dynamic-feed playbook Meta and Google have run for years. And measurement will feel like home too: there’s a pixel and a Conversions API so you can tie a ChatGPT impression back to an actual purchase, lead, or sign-up.
If your team already runs paid social, the learning curve here is short. That was true for us, and it’s the single biggest reason we think agencies and in-house teams can get productive fast.
The pricing reality (and the $3 question everyone asks)
Let’s clear up the number people keep asking about. There is no official minimum spend anymore. You can start a campaign on a modest daily budget and go.
But here’s the nuance we confirmed in testing. On CPC campaigns, OpenAI recommends a starting max bid of $3 to $5 per click, and in practice, bids under $3 tend to just not clear the delivery threshold. So while nobody at OpenAI calls it a “minimum,” $3 functions as the real-world floor for cost per click. Try to sneak in below it and your ads quietly stop serving.
A few things worth knowing about how you actually pay:
- The auction is a relevance-weighted, second-price model. Translation: a sharp, well-matched ad on a $3.50 bid can and does beat a lazy ad sitting on a $6 bid. Ad quality is not a nice-to-have here, it’s a cost lever.
- CPM buying still exists for reach-focused brand campaigns, with clearing rates that have softened well below the original launch pricing as more inventory came online.
- Conversion-based (CPA) bidding is rolling out for accounts with proper conversion tracking in place, which is where this gets genuinely interesting for performance budgets.
For a lot of B2B categories, a $3 to $5 click lands in the same neighborhood as branded and competitor terms on traditional search, and often cheaper. The catch, as always, is volume and intent quality, which is exactly the kind of thing you learn by running the channel, not reading about it.
The features we actually leaned on
The beta already ships with a solid toolkit:
- Guided campaign setup plus bulk upload for launching at scale
- Budget, objective, and pacing controls, including daily budgets
- Context hints, where you feed the system signals about the conversations you want to show up in
- Product feed ads for retailers
- Creative review and clear ad labeling
- Reporting with charts, aggregate totals across campaign, ad group, and ad views, and CSV export
- Pixel and Conversions API measurement
- Team access and permissions, so more than one person can run the account cleanly
- Dynamic calls-to-action that the system selects based on your creative and destination
It’s a beta, so expect formats, countries, and policies to keep shifting. We’d rather tell you that up front than pretend the thing is finished.
Where we come in
Here’s the honest part. The platform is approachable, but “approachable” and “profitable” are two different things. Getting into the account is easy now. Making a relevance-weighted auction inside a conversation actually pay back your spend is where the real work lives: writing ad copy that matches intent, structuring context hints, getting your pixel and Conversions API wired correctly so you’re not flying blind on attribution, and reading early signals before you scale.
That’s the work our US branch has been doing since day one of this beta. We’re already inside the platform, we’ve made the early mistakes so you don’t have to, and we run ChatGPT and OpenAI ad campaigns as a managed service for brands who want a seat at this table without building the muscle from scratch.
If you’re a US business and you’ve been curious whether ChatGPT ads are worth a test, this is the moment to find out, while the channel is still uncrowded and the cost of learning is low. We’re happy to walk you through what a first campaign would look like for your category, what to budget, and what “good” looks like this early.
Reach out and let’s map it. The conversations your customers are having with ChatGPT are already happening. The only question is whether your brand is in them.
