ChatGPT Ads Are Open to Everyone Now: A Small Team's Playbook for the New Self-Serve Ads Manager

What just changed

For most of the last year, advertising inside ChatGPT meant knowing someone at OpenAI. You needed a direct sales relationship, a decent budget, and patience. That gate is coming down.

OpenAI has opened its self-serve Ads Manager to US advertisers, which means you no longer need a sales rep to run a campaign. [1] That single change opens the platform to the long tail: mid-market brands, performance shops, and agencies testing on behalf of clients who aren’t ready to commit big sums yet. [2] If you run marketing for a small team, this is the first time the door has actually been open to you.

And the plumbing has filled in fast. In short order OpenAI has added conversion tracking, role-based account access, spend reporting at the ad level, daily budget controls, and cost-per-click bidding alongside its original cost-per-impression model. [3] Most recently it started testing richer ad formats: a larger image with an optional, customisable call-to-action button, plus a dedicated e-commerce format that pulls in price and customer reviews. [2]

Translation: ChatGPT ads went from “an experiment a few big brands could touch” to “a real platform a small team can log into.” That doesn’t mean you should pour your budget in tomorrow. But you should understand it now, because the brands learning it early are building an advantage that compounds.

Why ChatGPT ads are a different animal

Don’t think of this as another Google Ads or Meta Ads. The behaviour around the ad is fundamentally different, and that changes everything about what works.

When someone’s chatting with ChatGPT, they’re usually mid-task. They’re researching running shoes, planning a trip, comparing software, working a problem. It’s a high-intent, conversational environment, and the context of the query shapes what your ad actually needs to do. [2] An ad that lands while someone is researching marathon shoes has a very different job than one that lands while they’re planning a holiday.

That makes creative performance more variable than it is in a news feed or a search box, and far more sensitive to execution. As OpenAI’s vp of monetization Benji Shomair put it, “creative variation has been a real key to success.” [2] In plain terms: the same generic ad you run everywhere else will underperform here. Context-matched creative wins.

Should you move early? An honest pros and cons

Let’s be honest, most “be a first mover” advice is just hype dressed up as strategy. So here’s the real calculus.

Reasons to test now: the auction is less crowded than mature platforms, which usually means cheaper attention while it lasts. You learn a new channel before your competitors do. And the audience using ChatGPT skews toward exactly the kind of high-intent, considered purchases many businesses want.

Reasons to wait: the targeting is still thin. Audience targeting, the ability to retarget or build lookalikes, is only in a gated rollout. [2] Outcome-based optimisation, where you tell the platform to chase a cost-per-acquisition or return on ad spend instead of clicks, is still in development with no public timeline. [2] For performance marketers who live and die by CPA, that’s a genuine reason to hold serious budget back. And there’s a queue: even ad-tech insiders report waiting for account approval, with a message that verification “may take some time” due to high signup volume. [2]

My take: if you have a small test budget and the patience to learn, get in the queue now and run a modest experiment. If your marketing lives entirely on strict CPA targets, set a calendar reminder and revisit when outcome-based optimisation ships. Don’t bet the quarter on it either way.

The starter campaign: a step-by-step for a small team

Here’s how I’d run a first test without wasting money.

1. Sign up for the Ads Manager and expect a wait. Verification is backed up, so do this first while you build everything else. [2]
2. Pick one product with high intent. Not your whole catalogue. Choose the single offer where someone researching in ChatGPT is close to buying. High-consideration purchases work best here.
3. Install conversion tracking on day one. The pixel and conversion tools exist now, so use them from the start. [3,4] Running ads without measurement is just donating money.
4. Set a small daily budget and start on CPC. Use the daily budget control to cap risk, and start with cost-per-click so you only pay for actual engagement while you learn. [3]
5. Write three creative variations, not one. Because context drives performance here, build a few versions aimed at different mindsets (researching vs ready-to-buy) and let the data tell you which fits.
6. Give it two weeks and read the conversions, not the clicks. Clicks tell you the creative is interesting. Conversions tell you the channel is working. Judge on the second.

That’s a real test for a few hundred dollars, not a leap of faith.

The creative that actually fits

The format has evolved past the early days, when an ad was just a headline, a short description, an image, and a link. [2] Now you have more room to work with, so use it deliberately.

Use the dynamic CTA. The new format supports buttons like “shop now,” “book now,” “sign up,” and “learn more.” [2] Match the button to the intent. Someone deep in research wants “learn more,” not “buy now.” Pushing too hard too early kills conversational ads.

Lean on reviews in the e-commerce format. The dedicated shopping format pulls in price and customer reviews automatically. [2] If your product has strong ratings, this format does your selling for you. If your reviews are thin, fix that before you spend.

Write like a helpful answer, not a billboard. The ad sits inside a conversation. Copy that sounds like a useful recommendation outperforms copy that shouts. This is also where the lessons from getting recommended organically apply: we cover that in our playbook for getting your brand cited in ChatGPT. The same “be genuinely helpful” principle that earns organic mentions earns ad clicks too.

What’s still missing (so you plan around it)

I’d rather you go in clear-eyed than disappointed. Here’s what isn’t there yet.

• Mature targeting. Retargeting and lookalikes are gated; you can’t yet build the precise audiences you’re used to on Meta. [2]
• Outcome-based bidding. You can optimise toward clicks, not yet toward sales or leads directly. For many advertisers, that’s the feature that unlocks real budget, and it’s still in development. [2]
• Scale. This is early. The volume isn’t Google’s. Treat it as a learning channel, not a primary one, for now.

The pace of change, though, is aggressive, and OpenAI is reportedly preparing to file its IPO prospectus with a view to going public as early as September. [2] A company racing toward the public markets has every incentive to build out its ad business fast. The gaps you see today probably close faster than you’d expect, which is the real argument for learning the platform now rather than later. For the wider picture of where AI-driven advertising is heading, our recap of Google Marketing Live 2026 and our look at autonomous ad bidding are both worth your time.

A realistic word on budget and expectations

One thing I want to set straight, because I’ve watched teams burn goodwill over it: this is a learning budget, not a growth budget. Don’t walk into your boss’s office and promise ChatGPT ads will hit your quarterly revenue target. They won’t, not yet, and overpromising on an unproven channel is how good marketers lose credibility.

Frame it internally the way you’d frame any genuine test. You’re spending a small, capped amount to answer specific questions: does our audience engage here, which creative angle works, and what does a tracked conversion actually cost? Those answers are the return on this spend, even if the direct sales are modest. The brands that will dominate ChatGPT ads in a year are the ones building that knowledge now, while the auction is cheap and the competition is light.

Set the expectation low, measure honestly, and let the learning compound. That’s a far stronger position than a splashy launch you can’t sustain when the targeting and optimisation features finally arrive.

What to do this week

Three moves. First, sign up for the Ads Manager today so your account is in the verification queue while you prepare. Second, pick the single product where a ChatGPT user is closest to buying, and write three creative variations for it. Third, make sure conversion tracking is ready to go so your first test produces real data, not vibes.

You don’t need to bet big. You need to be in the room early, learning a channel that’s clearly going to matter, while it’s still cheap to make mistakes.