ChatGPT Ads Just Went Global: The Agency Playbook for a $1 Billion Channel

On October 21, OpenAI pushed ChatGPT ads into 26 more countries, bringing the total to 89. Fourteen of the new markets are in Latin America — Argentina, Chile, Colombia, Peru — plus Nigeria, South Africa, Serbia, Ukraine, Cambodia, Bangladesh, and Pakistan. The company says more than 100,000 advertisers have already run campaigns, and it reported a $1 billion annualized revenue run rate back in August. For context, that’s a billion-dollar ad channel built in about eight months, starting from a US-only test earlier this year.

Ads still only run on the Free and Go plans; Plus, Pro, and Enterprise stay ad-free. Advertisers get in through tech partners or a self-serve Ads Manager.

Here’s why this matters to you as an agency owner rather than just a marketer. New ad inventory follows the same script every time. Early on, attention is cheap and expertise is rare. A year from now, attention is expensive and every agency offers the service. The ones that sell the transition work early get the clients; the ones that wait sell nothing. We watched this exact movie with mobile, with Facebook Shops, and with TikTok Shop — the agencies that moved early sold the transition work, and the agencies that waited competed on price for scraps.

Your clients have already heard the headline. They cannot describe what they want from ChatGPT ads — they’ll just ask if they should be on it. That’s the perfect buying posture: vague anxiety instead of a sharp brief. Clients with a sharp brief price-shop you. Clients with vague anxiety pay for guidance. We covered the same dynamic when we wrote about selling AI shopping-agent readiness as a service, and the playbook rhymes exactly.

What the channel actually gives you to sell

Forget the hype for a minute and look at the infrastructure, because infrastructure is what turns an experiment into something you can bill against. OpenAI has spent the last few months building the measurement stack first: the ChatGPT Ads Pixel and Conversions API track purchases, leads, installs, and sign-ups. The attribution list runs through AppsFlyer, Triple Whale, Adjust, Branch, Singular, Kochava, Airbridge, and Tenjin. Fospha, Measured, and INCRMNTAL handle full-funnel measurement; Haus, Measured, and WorkMagic run geo-based incrementality tests. Brand-suitability pilots with DoubleVerify and Integral Ad Science check ad environments without touching private conversations, and “Negative Phrases” controls let qualifying advertisers define placement exclusions. There are also AI-powered creative tools for advertisers, rolled out in September, and a “business-sponsored agents” test running with select US advertisers.

This is the boring plumbing agencies usually get paid to install. The mistake we see most often with a new channel has nothing to do with creative — brands spend the first month arguing about headlines while their tracking is half-wired, then spend the second month not trusting their own numbers. Sell the wiring first. A pixel-and-API setup engagement is the easiest yes you’ll ever get, because every client on the platform needs it and almost none of them have it right on day one. ChatGPT’s visual ad format launched earlier this month with the same measurement stack behind it — the advertisers who wired tracking before the test went live were the only ones who could read their own results.

What we’d actually put in the proposal

Do not sell a one-off “ChatGPT ads audit” PDF. The audit is the vaguest work in this category: a 30-page deck with a red-yellow-green table, delivered once, invoiced once, never opened again. Audits answer a question the client stops caring about the day after they read it. Sell ongoing work instead — and the difference between a proposal that gets a reply and one that gets ignored is exactly this level of concrete detail.

Here’s the package we’d pitch this month:

1. Conversational-readiness review. Not every brand belongs in a chat surface. Products people research, compare, and ask questions about — home goods, fashion, health supplements, B2B tools — fit naturally. Impulse buys and coupon-driven offers don’t. Tell the client honestly which one they are. Saying “you shouldn’t advertise here yet” to the wrong client is the fastest credibility-builder in the room, and it makes the yes from the right client easy.

2. Creative built for the surface. The lazy move is uploading the same static product banner you run on retargeting. It dies in a creation flow. ChatGPT users are mid-question, mid-imagination — the creative that works reads as inspiration or a useful answer, not an interruption. We’d budget three to five variants per client to start, because nobody knows which visual register the placement rewards yet. Brands with real usage photography are halfway there. Brands with white-background packshots and discount codes need a shoot before they spend a dollar.

3. Measurement before media. Pixel, Conversions API, one attribution partner the client already uses, and a baseline from their current channels. Wire it in week one. This is also the part that justifies a retainer: attribution stays honest only if someone keeps watching it.

4. A monthly management retainer with a test budget cap. Price it as a starter service — a modest monthly fee plus a capped test budget the client controls. The pitch that works: “We’ll get you live on a new billion-dollar channel, wired and measured, for less than the cost of one wasted week of Meta spend.” You’re selling learning-curve arbitrage. The auction is still cheap because 100,000 advertisers is tiny next to the millions bidding on Meta.

The bad pitch to avoid

We keep seeing the same failure pattern when agencies chase a new channel: they lead with the novelty, promise “first-mover advantage,” and set expectations the platform can’t cash. ChatGPT ads are on Free and Go plans only — the most casual segment of the audience, people who won’t pay $20 a month for the product. Nobody has proven that audience converts for high-ticket offers at scale. If your client’s average order value is $300 and you promise pipeline from a free-tier chat surface, you’re setting up a painful month-three review call.

Lead with the experiment framing instead. “This is a learning campaign, not a scale campaign” is the sentence that saves the relationship. Test budgets stay small, timelines stay honest, and when something does convert, you look like the genius who found it early instead of the agency that overpromised.

The honest caveat

We’re not sure how fast this normalizes. A $1 billion run rate in eight months means the auction is already maturing — the cheapest days are probably behind us, even if they’re cheaper than Meta’s mature auctions. OpenAI is reportedly heading toward an IPO, and ad inventory is now doing real work on the revenue story, which means pressure to open more placements and possibly loosen the Free-and-Go boundary over time. That’s good for scale, bad for the early-mover discount.

So treat this as a real but bounded opportunity: get clients in, get measurement wired, learn the creative grammar of the surface — and don’t reroute money from channels that are already working just to be early on one that might stay experimental for another year. The agencies that win this channel will be the ones with data from the first wave, not the biggest day-one budgets.

FAQ

Which ChatGPT plans actually show ads?

Free and Go only. Plus, Pro, and Enterprise stay ad-free, which is exactly why the audience here skews casual — and why promising high-ticket pipeline from this surface is a fast way to lose a client in month three.

What should an agency sell first on this channel?

The wiring, not the creative. Pixel plus Conversions API, one attribution partner the client already uses, and a baseline from their current channels. Every client needs it, almost none of them have it right on day one, and it’s the easiest yes in the whole package.

Is this a scale channel or a test channel right now?

A test channel with a maturing auction. A $1 billion run rate built in about eight months means the cheapest days are probably already behind us, even if the auction is still cheaper than Meta’s. Sell learning-curve arbitrage, not scale.

Which clients actually fit ChatGPT ads?

Products people research, compare, and ask questions about — home goods, fashion, health supplements, B2B tools. Impulse buys and coupon-driven offers don’t fit a surface where the user is mid-question, and telling a wrong-fit client “not yet” is the fastest credibility-builder in the room.