StartOn January 20, Sam Altman announced that ChatGPT would start testing ads, reversing a position he'd held publicly for years. On February 9, ads went live for free and Go-tier users in the US. By March, they'd already crossed $100 million in annualized revenue. By May, OpenAI had built a full ad-tech stack, cost-per-action billing, a self-serve ads manager, conversion-optimized campaigns, live across nine countries. In the weeks immediately after that buildout, OpenAI confidentially filed to go public.
Calling this a test was accurate in February. It stopped being accurate somewhere around March, and the timing of everything that followed is worth laying out in order, because the ad format itself is also genuinely new, not just a faster version of something that already existed.
A New Kind of Ad
Every prior generation of digital advertising, banner, search, social feed, shows you something and waits for you to leave the page to act on it. OpenAI's own description of what it's building is different in a specific, structural way: with a conversational interface, the company wrote, "you might see an ad and be able to directly ask the questions you need to make a purchase decision." Picture the difference between a billboard and a salesperson standing next to it who can actually answer whatever you ask on the spot. A traditional ad ends the moment you look away. A ChatGPT ad can keep the conversation going inside the same window, with the same model that was just helping you with something else now fielding follow-up questions about a product.
The mechanics as OpenAI has described them: ads appear below a response, clearly labeled as sponsored, visually separate from ChatGPT's actual answer. They're limited to logged-in adult users on the Free tier and the $8-a-month Go plan. Plus, Pro, Business, Enterprise, and Edu accounts never see them. Targeting defaults to conversation topic, chat history, and past interactions with ads, and OpenAI says it won't place ads near health, mental health, or political topics, won't show them to anyone flagged as under 18, and won't let advertisers pay to influence what ChatGPT actually says. Free users can turn ads off entirely, but the price isn't money, it's a lower daily message limit. Ad-free costs you usage instead of cash if you're not already paying.
None of those specific promises have been independently verified by anyone outside the company. They're the stated rules, not an audited outcome, and that's worth being precise about given how much of what follows depends on people trusting a system that's reading their conversations to decide what to show them.
The Part That Actually Reached People
Everything up to this point was covered in a company blog post, which is exactly the kind of thing most users never read. What actually put this in front of people was an email, and I got one myself this week, addressed to me directly. Here's what it said, in full:
We're updating our Privacy Policy to include information about ads in ChatGPT, including how they work and how you can control your experience.
Ads may appear on Free and Go plans. Plus, Pro, Enterprise, Business and Education plans do not have ads. Ads do not influence the answers ChatGPT gives you. When you see an ad, it is always clearly labeled as an ad and is visually separated from the organic answer. You'll get relevant and personalized ads using information that stays only on ChatGPT, such as ads you've interacted with, or context from your chats. You can manage ad personalization anytime in settings.
Your personal details and conversations with ChatGPT are private and are not shared with advertisers. Advertisers do not have access to your chats, chat history, memories, or personal details. Advertisers only receive overall information about how their ads perform, such as total views or clicks.
Reading it as the actual recipient rather than as a report about it changes what stands out. It's short, plainly worded, and reads more like a routine terms-of-service update than the announcement of a new advertising business, which is presumably deliberate. It also arrived without any indication that this was new or notable, no fanfare, just a policy update landing in an inbox alongside whatever else was there that day.
I'm in India, and the email arrived as OpenAI's ads program was reaching Indian users. But it was not the first public notice of an international rollout. OpenAI had already announced the sequence: on March 26 it said the pilot would expand beyond the US into Canada, Australia, and New Zealand; on May 7 it said the next wave would cover the UK, Mexico, Brazil, Japan, and South Korea. India was not named in those announcements, but the broader direction was public long before this email. The message I received was therefore a direct privacy-policy notice to an existing user, not a substitute for a rollout announcement. What began as a US-only pilot in February had become a staged international expansion within six months, with public announcements setting the sequence and direct emails explaining what the change meant for users in each market.
The specifics inside it are worth stating plainly too. Starting in February 2026, if Personalized Ads is enabled, targeting can draw on the current chat thread itself, including how the model responded in it, not just topic-level signals. Advertisers still don't receive chat content, chat history, memories, names, emails, precise location, or sensitive topics like health or political views, they get aggregated performance numbers like views and clicks, consistent with what my own email said. That same privacy policy update also introduced an opt-in contact-syncing feature elsewhere, letting a user connect their device address book so OpenAI could identify which contacts also use its products. Privacy researchers flagged a specific wrinkle: even a user who never enables the feature themselves could still have their own phone number processed if someone else in their contacts synced theirs. OpenAI has since walked back parts of that feature after the backlash, but it shipped in the same broader update as the ads notice, and it's a large part of why this rollout, not the original January announcement, is what actually generated public unease once people started reading past the first paragraph.
It's also worth being precise about what Sam Altman actually said before reversing course. In 2024, he'd described advertising as a last resort for the company, not a plan, an option to reach for if nothing else worked. Fourteen months later, it's the fastest-growing new product line the company has, expanding into new countries roughly as fast as the emails announcing it can go out.
The Ramp Has No Real Precedent
What makes this different from a normal ad rollout is the speed. Six weeks from a standing start to $100 million in annualized revenue is a pace that doesn't have a clean comparison in prior consumer tech advertising launches, most ad products take years to get there, not one fiscal quarter. By May, OpenAI wasn't just running ads, it had built the infrastructure of a mature ad platform: self-serve buying tools any advertiser could use directly, cost-per-action pricing that lets marketers pay for actual outcomes instead of impressions, and a footprint across nine countries. That's the difference between piloting a feature and standing up a business unit, and OpenAI did the second thing in about four months.
Who's actually seeing these ads matters for reading that revenue number correctly. Reporting on the early advertiser base describes the audience skewing toward individual contributors, founders at small companies, students, and researchers, people using personal accounts for their own work rather than employer-funded seats. The senior, enterprise-funded users most likely to influence a big B2B purchase are disproportionately on the ad-free paid tiers already. What OpenAI built in six months is a genuinely large-scale consumer ad business sitting on top of hundreds of millions of free-tier users, not a B2B lead-generation tool.
The Two Weeks Before the Filing
Here's where the timing gets specific enough to be worth stating plainly rather than implying. The cost-per-action billing, the self-serve manager, and the redesigned ad unit all landed in a tight window in May, and in the two weeks immediately following that buildout, OpenAI submitted a confidential draft S-1 to the Securities and Exchange Commission. That filing came on June 8, exactly one week after Anthropic filed its own confidential S-1 on June 1, and days before Elon Musk's SpaceX, which had absorbed xAI into its structure earlier in the year, began its own IPO roadshow.
OpenAI's own announcement was characteristically blunt about why it was going public with the news at all: "We recently submitted a confidential S-1. We expect it to leak so we're just announcing it." The company said timing remains undecided, early speculation pointed to a listing as soon as the fourth quarter of 2026, though more recent reporting suggests CFO Sarah Friar has told associates 2027 is the more realistic internal target. Valuation at the time of filing sat around $850 billion, with some analysts modeling a public debut north of $1 trillion.
None of this proves the ad rollout was deliberately timed to strengthen a specific filing date. It's also not a coincidence worth ignoring: a company turned a two-month-old advertising experiment into a real, measurable, self-serve revenue product in the same stretch it was preparing the paperwork that would expose its full financials to public markets scrutiny for the first time.
Why the Timing Actually Matters
OpenAI is, by its own CFO's account, not yet profitable, generating roughly $2 billion a month in revenue that's reportedly growing about four times faster than Alphabet's or Meta's did at a comparable stage, with enterprise customers now making up more than 40% of that revenue and expected to reach parity with consumer revenue by the end of the year. That's a strong growth story built almost entirely on subscriptions and API access. What it was missing was a second, genuinely large consumer revenue line that didn't depend on someone deciding to pay $20 or $200 a month.
Ads solve exactly that gap, and they solve it fastest by monetizing the hundreds of millions of people who were never going to become paying subscribers in the first place. For a company about to ask public investors to underwrite a valuation approaching a trillion dollars while still burning cash on compute, walking into that conversation with a new, fast-growing, genuinely novel revenue category already live and generating real numbers is a materially different pitch than walking in with a subscription business alone.
Ads in ChatGPT stopped being a test the moment they had their own billing system, their own advertiser tooling, and a growth curve fast enough to matter to a trillion-dollar filing. The format itself, an ad you can interrogate instead of one you scroll past, is new enough that nobody else has run it at this scale yet. Given how quickly OpenAI built the business around it, it likely won't stay unique for long.
Previously on TheQuery: