Seven months.
That’s roughly how long it took ChatGPT to go from testing ads in the U.S. to building an advertising business running at $1 billion in annualized revenue.
For perspective, I’ve seen marketing teams take longer than that to agree on a new homepage.
According to OpenAI, ChatGPT Ads reached that $1 billion annualized run rate in fewer than 200 days. Tens of thousands of advertisers are already using the platform, and its advertising footprint continues to expand globally.
There’s just one slightly awkward question: Do the ads actually work?
We don’t really know yet. And that’s what makes this story so interesting.
First, That $1 Billion Needs an Asterisk
ChatGPT hasn’t collected $1 billion in advertising revenue. It has reached a $1 billion annualized revenue run rate, meaning its current revenue pace, extended across a full year, gets you to roughly $1 billion.
Still impressive, especially considering OpenAI only began testing advertising in ChatGPT in February 2026.
By late August, the ad business had crossed the billion-dollar run-rate threshold. Axios reported that WPP Media considers generative AI search advertising the fastest-growing ad format in history.
So naturally, marketers want in. We do love a new channel.
The Pitch Is Almost Irresistible
Imagine you’re planning a vacation with ChatGPT. You tell it there are four of you, you’re traveling over spring break, you want somewhere warm with a great pool and a direct flight, and you’d like to keep the whole thing under $6,000.
That’s very different from typing “Mexico vacation” into a search box.
You’re not just supplying keywords. You’re explaining what you’re actually trying to accomplish, including your budget, preferences, constraints and probably a few things you absolutely do not want.
That distinction matters.
OpenAI says its advertising system is designed to show clearly labeled sponsored content separately from ChatGPT’s answers, while keeping conversations private from advertisers.
For marketers, this is catnip because we’ve spent years trying to piece together breadcrumbs of intent. What did someone search? What did they click? What did they watch? What did they put in their cart and abandon while reconsidering every financial decision they’ve made since college?
Conversational AI potentially puts advertising much closer to the moment someone is actually working through a decision.
And apparently OpenAI has decided even that isn’t close enough.
And Now the Ad Can Talk Back
This week, OpenAI announced Sponsored Agents, a new format being tested with select U.S. advertisers.
Instead of clicking an ad and immediately being sent somewhere else, a consumer can choose to start a separate, clearly labeled conversation with an AI agent sponsored by the advertiser.
Think about shopping for a dining table. You see a sponsored placement, but rather than immediately hitting Shop Now, you ask whether the table will fit in your 10-by-12 dining room, whether six people can sit comfortably around it and whether the finish is going to survive two kids and a Labrador.
Those aren't examples I dreamed up. OpenAI uses the dining-table scenario to explain the concept. The consumer can ask follow-up questions inside the sponsored conversation and visit the advertiser’s website when they’re ready. OpenAI says that sponsored conversation remains separate from the original ChatGPT conversation and from ChatGPT’s independent answers.
It sounds like a product feature, but I think the marketing implications are considerably bigger.
The Landing Page Just Got a Competitor
For roughly 25 years, digital advertising has been built around getting someone to leave whatever they’re doing. We show an ad, hope for a click, send them to a landing page, make them navigate around looking for information and then hope they convert.
We’ve spent an enormous amount of time optimizing that journey: page speed, navigation, forms, CTAs, personalization and, of course, A/B tests of button colors that somehow required 14 people and three meetings.
Sponsored Agents could collapse part of that journey into the advertisement itself.
Instead of digging through a product page, FAQ section, reviews and shipping policies, a consumer could simply ask the brand what they want to know. That doesn't make the website irrelevant, but it could change where persuasion happens.
OpenAI is also beginning to connect the infrastructure around that experience. The company announced HubSpot as its first CRM partner and Shopify as its first ecommerce partner. Advertisers can use natural-language prompts to create, update and analyze campaigns, while optional AI tools can adapt existing ad copy to better fit the context of a conversation.
Suddenly an ad isn’t simply creative. It’s part media placement, part product expert, part salesperson and part customer experience.
Which creates another fun question for the org chart: Does marketing own it? Sales? Ecommerce? Customer experience? IT?
I’m sure we’ll invent a new acronym and a VP title for it by Christmas.
Then the CFO Walks Into the Room
This is where the billion-dollar story becomes much more interesting, because whether the ad is a sponsored placement or an AI agent capable of having a conversation, somebody eventually asks the same question:
Cool. What did it sell?
The Wall Street Journal reported that OpenAI’s extraordinarily fast advertising buildout has also produced some growing pains. Early advertisers have encountered limited targeting options, operational glitches and measurement challenges.
Rocket CMO Jonathan Mildenhall described his company’s initial performance as “underwhelming” and said he hadn’t yet seen a material impact from testing. Taskrabbit has continued experimenting despite some early administrative headaches but wants better targeting and more information about how consumers search inside ChatGPT before increasing its investment.
Hungryroot, meanwhile, has seen more encouraging early results. Its CMO pointed to something particularly interesting: chatbot queries can be much more detailed than traditional searches, potentially giving advertisers a richer picture of what someone is actually trying to accomplish.
That’s probably the most accurate description of where we are right now. Some marketers see real promise, some aren’t impressed yet, and almost everybody wants to test it.
We’ve Seen This Movie Before
Every major digital advertising platform goes through some version of this. First comes the audience, then the advertisers, then the case studies, and eventually somebody in finance asks why we’re spending $2 million there.
Facebook went through it. Streaming went through it. Retail media went through it. Influencer marketing is still going through it.
Now it’s AI’s turn.
The difference is that conversational advertising creates a particularly interesting attribution problem. Imagine someone spends 20 minutes talking with ChatGPT about running shoes. They compare four brands, see a sponsored placement, open a Sponsored Agent and ask three more questions. Two days later, they walk into a store and buy the shoes.
What deserves the credit? The original ChatGPT conversation? The sponsored placement? The agent interaction? The retailer? The Google search they did the night before?
Welcome back, attribution. We missed you terribly.
Measurement Has to Catch Up
OpenAI clearly understands this. The company has been building its advertising infrastructure at remarkable speed, but independent measurement will become increasingly important as larger advertisers start moving meaningful budgets onto the platform.
In an interview with Digiday, OpenAI’s David Dugan described third-party measurement as a “natural step” as the platform matures. He also acknowledged that measuring advertising inside conversational AI may require a different playbook from traditional search.
That makes sense because a click is easy to count. A conversation is much harder.
If someone asks a brand’s Sponsored Agent six questions, spends eight minutes interacting with it, never clicks through and purchases somewhere else three days later, that interaction probably had value. Figuring out exactly how much value is where things get messy.
And now we’ve got ourselves a marketing meeting.
The CMO Lesson
I absolutely think marketers should be experimenting here. A platform that has already attracted tens of thousands of advertisers and built a billion-dollar annualized advertising business in fewer than 200 days deserves attention.
But there’s a difference between being early and being reckless.
The smart move is to test, learn how consumers actually behave, understand whether conversational intent produces better customers and start building the measurement infrastructure before the experimental budget suddenly becomes a very real line item on next year’s media plan.
Sponsored Agents make that work even more important because we're no longer just talking about another place to put an ad. We may be watching the advertisement itself evolve into part of the customer journey.
That’s a significant shift.
ChatGPT built a $1 billion advertising business in less than 200 days, and now the ads themselves can start conversations. The technology is moving extraordinarily fast, but the fundamental marketing question hasn’t changed:
Did it work?
The future of advertising may very well be conversational. The CFO is still going to want the spreadsheet.
