OpenAI Just Opened ChatGPT Ads to Everyone. The Channel That Didn't Exist Four Months Ago Now Has 600 Advertisers, $100 Million in Revenue, and a Self-Serve Platform.

We recently wrote about the moment OpenAI started testing ads inside ChatGPT. The piece was called "OpenAI Put Ads in ChatGPT. The Trust Contract With AI Search Is Already Over." The argument was simple: the thing that made AI search valuable to users was the perception that it was neutral. A layer between them and information that didn't have a financial stake in the answer. Ads changed that permanently. It doesn't matter where they sit on the page or how clearly they're labeled. The relationship shifted.

On May 5, OpenAI launched a beta self-serve Ads Manager at ads.openai.com. The $50,000 minimum spend that gated the original pilot is gone. Any U.S. advertiser can now register, upload creative, set budgets, control pacing, and track campaign performance directly through OpenAI's platform. No agency required. No managed relationship. No gatekeeper standing between a small business and an ad placement inside the tool that 900 million people use every week for advice.

That last part is what makes this different from every other ad platform launch in the last decade. And it happened with very little coverage relative to how big the implications actually are.

$100 Million in Six Weeks

The original ChatGPT ads pilot launched on February 9 with a managed group of large advertisers: Best Buy, Lowe's, Target, VistaPrint, Albertsons, Williams-Sonoma. Within six weeks, the pilot crossed $100 million in annualized revenue, according to a company spokesperson who shared the figure with Reuters on March 26.

Six weeks. $100 million annualized. With fewer than 20% of eligible users seeing ads on any given day.

OpenAI now has more than 600 advertisers on the platform. The company told investors it expects $2.5 billion in ad revenue by the end of 2026. By 2029, it projects $53 billion. By 2030, $100 billion. Those projections assume ChatGPT reaches 2.75 billion weekly active users by 2030, up from roughly 900 million today. There is a lot of room to question whether those numbers hold. But the trajectory from pilot to nine figures in under two months is not a projection. That already happened.

Here is the math that explains why this was always coming. ChatGPT has around 900 million weekly active users, but only about 59 million pay. That breaks down to roughly 50 million consumer subscribers and 9 million business users. The conversion rate sits at 6.5%. When more than 90% of your user base is a cost center, you need another revenue model. Sam Altman said in 2024 that he thought of ads as "a last resort." He hired Shivakumar Venkataraman from Google Ads to build the infrastructure. He hired David Dugan from Meta to run global advertising solutions. He put Asad Awan in charge of monetization. That is not a "last resort" hiring strategy. That is a deliberate recruiting raid on the two companies that built modern digital advertising.

The "last resort" arrived on schedule.

How It Works

ChatGPT ads show up as labeled sponsored boxes at the bottom of AI responses. They include the advertiser's name, favicon, headline, copy, an image, and a landing page link. They sit below the organic answer, visually separated. OpenAI calls its governing principle "Answer Independence," meaning advertising relationships do not influence what ChatGPT actually says.

That is the stated position. Whether users continue to believe it is the question that will determine whether this channel scales or collapses. More on that below.

Targeting does not work like Google or Meta. There are no keyword lists. No demographic segments in any traditional sense. ChatGPT matches ads to conversations using what OpenAI calls "context hints." These are broad thematic descriptions of the questions and needs users bring to the platform. The system reads the current conversation, the user's profile and preferences if they're logged in, and their previous ad interactions. No user data goes to advertisers. No chat transcripts. No names. No emails. No IP addresses. No precise locations. All reporting is aggregated.

Buying now supports both CPM and CPC. The initial pilot was CPM only, with a default max bid around $60. The May 5 expansion added cost-per-click at a recommended starting bid of $3 to $5 per click. OpenAI also shipped a Conversions API and pixel-based measurement for tracking post-click actions: purchases, leads, sign-ups.

For performance marketers, CPC plus conversion tracking makes ChatGPT feel like a real media channel instead of an experiment. That is exactly what OpenAI intended.

Campaigns follow a three-level structure: campaign, ad group, ad. Every campaign goes through review before it serves. Ads currently reach users on the Free and Go tiers in the U.S., Canada, Australia, and New Zealand. Paid subscribers on Plus, Pro, Business, Enterprise, and Education do not see ads. No one under 18 sees ads. No ads appear near politics, health, or mental health topics.

Agency partners: Dentsu, Omnicom, Publicis, WPP. Technology partners: Adobe, Criteo, Kargo, Pacvue, StackAdapt. International expansion to the U.K., Brazil, Japan, South Korea, and Mexico is expected in the coming weeks.

The Signal Is Different and That Is the Whole Point

I keep thinking about this one detail because it explains why the ad industry is paying attention even though the click-through rates are, frankly, bad.

Most ad platforms launch into environments where the user behavior is well understood. You know what people do on Instagram. You know what a Google search query signals. A media buyer's job is to match creative and targeting to known patterns.

ChatGPT breaks that model. When someone types "which CRM should I buy for a 50-person sales team," they have already disclosed budget context, team size, use case, and active purchase intent in a single exchange. Compare that to a Google search for "best CRM software," where the advertiser infers intent from a few keywords and hopes the landing page does the heavy lifting. In a ChatGPT session, the user told you what they need, why they need it, and what constraints they're working within before the ad even loads.

That is a richer commercial signal than anything search has ever offered. It is the reason the big agencies signed on early and the reason OpenAI can credibly project billions in revenue against a user base that mostly isn't paying for the product.

But. The engagement numbers are still rough. Advertisers are reporting click-through rates around 0.91%, per Adweek. Google search benchmarks at roughly 6.4%. That gap is enormous. The format is new. The placement is still being iterated. User expectations inside a conversational AI session are different from what they expect on a search results page, and it is going to take time for both sides to figure out what works.

Awan has been open about this. He said CPC bidding was introduced specifically to help prove ROI and give advertisers a reason to keep testing. That is the kind of transparency you want from a platform in its first year. Whether it lasts when the revenue pressure intensifies is another question.

The honest take: ChatGPT ads are a high-signal, low-maturity channel right now. The intent data is the best in the market. The infrastructure is early. The measurement is basic but moving fast. And the tolerance users have for ads inside what was sold to them as a neutral AI assistant is still an open question with a very short track record.

The Trust Problem We Wrote About Has Not Gone Away

This is the section that matters most to brand leaders. It is also the section that gets buried in every trade press writeup about the platform mechanics.

When we covered the initial launch, the core argument was about the trust contract between AI platforms and their users. People adopted ChatGPT because they believed it was giving them answers with no financial agenda. Ads change that dynamic even if the underlying architecture keeps them separate. Perception is what matters here, not plumbing.

OpenAI's own data says there has been "no impact on consumer trust metrics" and that ad dismissal rates are low. Fewer than 7% of ads get rated by users as "low relevance." Those numbers look good. They are also the company's own numbers, reported three months into a pilot with very light ad load, where fewer than one in five eligible users sees an ad on any given day.

The real test comes when the load increases. When the ad categories expand beyond consumer goods and travel. When the format starts evolving beyond a box at the bottom of the page. When users who are asking about medications or financial products start noticing sponsored recommendations in the same window.

Anthropic built its entire Super Bowl campaign around this tension. The ad positioned Claude as the AI that will stay ad-free. It was a direct shot at OpenAI and it landed because it said out loud what a lot of users were already thinking. This concern is not theoretical. It is the single biggest variable in whether ChatGPT ads become a durable channel or a short-lived cash grab that erodes the platform's core value.

We have covered this dynamic from multiple angles at this publication. In "Meta Just Automated Your Ad Strategy. The Only Thing Left to Control is Brand," we wrote about what happens when platforms absorb the execution layer and brands lose control of everything except their own identity. The same logic applies here, only the stakes are higher because the surface area is a conversation, not a feed. Your ad is not sitting next to content. It is sitting next to what users perceive as a direct, personal answer to a direct, personal question.

For brands advertising on ChatGPT, the question is not only whether the ads perform in Q3. It is whether the platform your ad sits on maintains the trust that gives the placement its value. An ad inside a trusted AI assistant is worth real money. An ad inside a platform that users are starting to view as commercially compromised is worth significantly less. Those two outcomes are both on the table right now, and which one materializes depends on decisions OpenAI makes over the next year, not on anything the advertiser controls.

What to Do With This

Test it. But go in with a clear understanding of what you are testing and what you are not.

ChatGPT accounts for 82.6% of all generative AI traffic. Comscore data from January showed 72.9 million unique U.S. visitors averaging nearly 8 minutes per visit. The sessions are active. Users are comparing products, evaluating services, asking for specific recommendations before making purchases. That is valuable attention and it is concentrated in a way that social feeds and display networks cannot replicate.

The self-serve platform removes the access barrier. CPC bidding gives you a measurement framework. The minimum spend is gone. You can run a meaningful test for the cost of a small paid social experiment.

Categories currently open: household and consumer goods, local services, travel and entertainment, digital products and education. Healthcare, finance, and several other regulated categories are still restricted.

If you run a B2B SaaS company, a professional services firm, an education business, or a high-consideration e-commerce brand, this channel deserves a line in your Q3 media plan. The intent signals are strong. Competition for inventory is still low. Early advertisers are locking in lower CPCs and building optimization data before the rest of the market catches up.

If your brand depends on being perceived as trustworthy, neutral, or independent, think carefully about the adjacency. Your ad is appearing inside a tool that people treat like a personal advisor. The context shapes how they perceive you. Make sure the creative earns its place there.

And whether or not you choose to run ads on ChatGPT, understand that this channel exists now and it is scaling fast. It changes how consumers discover products. It changes the competitive landscape for search. It forces every brand to treat AI as a discovery surface, not just a back-office productivity tool.

We wrote in February that the trust contract with AI search was already over. The self-serve expansion confirms it. The next chapter is about who builds inside that reality and who sits it out.