The scariest chart in advertising right now isn't a chart. It's a chat window.
Less than 200 days after OpenAI flipped the switch on ads inside ChatGPT, the company is already past a $1 billion annualized revenue run rate. It got to its first $100 million in annual recurring revenue in under two months. It's working with more than 600 advertisers, and it did it all while deliberately showing ads to fewer than 20% of eligible users on any given day.
Let that last part sink in, because it's the whole story. ChatGPT Ads is a $1 billion business running at roughly one-fifth throttle. If OpenAI simply stopped being cautious, the triopoly — Google, Meta, Amazon — that has vacuumed up more than half of the world's digital ad spending would be looking at a fourth entrant with the one thing they can't buy: the most high-intent traffic on the internet, sitting inside a conversation.
The moomoo piece that kicked this off asked the right question — should Google, Meta and Amazon be worried? — but it skimmed over the mechanics. So here's the deeper dig: how OpenAI actually makes money from ads, who buys them, how a ChatGPT ad gets rendered, what the exact same ad costs on the other platforms, and how a $1 billion run rate stacks up against $500 billion of incumbent revenue.
There are exactly two ways to buy a ChatGPT ad right now, and they serve two very different kinds of advertisers.
Door one: the ad-tech partner path. OpenAI is running its commercial engine through the platforms advertisers already use. Agency partners include the big four holding companies — Dentsu, Omnicom, Publicis, and WPP — while technology partners include Adobe, Criteo, Kargo, Pacvue, and StackAdapt. This is the commerce-heavy door: Criteo and Pacvue pull structured product feeds straight from the same catalog files a retailer already sends to Google Shopping, and the system auto-generates sponsored-product ads. The partner handles budgeting, bidding, and creative. OpenAI's system controls every delivery decision.
Door two: the self-serve Ads Manager. As of May 5, 2026, any U.S. business with a credit card can sign up, set a budget, upload a "chat_card" ad, and launch. There is no minimum spend — a meaningful change from the early pilot, where a test required roughly a $200K managed-service commitment. That's how a channel goes from "invite-only experiment" to "open marketplace" in a matter of weeks.
The buying models are intentionally simple: CPM (pay per thousand impressions) for reach, and — the important new addition — CPC (pay per click) for performance. OpenAI's stated logic for CPC is worth reading twice: ChatGPT conversations are "active and decision-oriented," so a click inside one is a meaningful signal that an ad was relevant. The company is also rolling out a Conversions API and pixel-based measurement, so advertisers can see what happens after the click — a purchase, a sign-up, a lead — while OpenAI keeps the actual conversation private.
This is where ChatGPT diverges from every platform that came before it, and where the "should Google worry" question gets its teeth.
A ChatGPT ad is a single unit OpenAI calls a chat_card: a 3–50 character title, up to 100 characters of body copy, one image, a favicon, and a destination URL. No video. No carousel. No rich interactive unit — yet. It appears at the bottom of the AI's answer, clearly labeled as sponsored, and it never sits inside the response itself. OpenAI's stated principles are blunt: the answers stay independent, the conversation stays private, and the ad never influences what the model says.
Targeting is contextual, not keyword-based. An advertiser doesn't bid on "best running shoes 2026." Instead, they provide broad "context hints" — descriptions of the situations a relevant customer might bring to ChatGPT — and the system matches those hints against conversation topics, prior chat history, and past ad interactions. This is a fundamental break from Google's search auction, where intent is captured in the query string itself.
The guardrails are unusually strict for an ad product. Users under 18 never see ads. Ads are withheld near sensitive topics like politics, health, and mental health. And roughly 85% of U.S. free and Go-tier users are eligible to see ads, but fewer than 20% actually do on a daily basis. OpenAI is running this conservatively on purpose — it's protecting the trust metric that makes ChatGPT valuable in the first place, and it has said the slow rollout has produced no impact on privacy-related trust.
The strategic point nobody's saying loudly enough: contextual matching on conversational intent is a different — and arguably deeper — signal than a search query. Someone typing "hotels under $200 in Tokyo" into Google has intent. Someone telling ChatGPT "I'm planning a Tokyo trip with a $2,000 budget, I like quiet neighborhoods, walkable, and I hate big chains" has specified intent. If OpenAI can monetize that second moment, it isn't just competing for search budgets — it's competing for the highest-value slice of them.
Here's the pricing reality, side by side. Benchmark figures are from 2025–2026 industry data (sources below), and the ChatGPT numbers are from OpenAI's own guidance and advertiser guides.
| Platform | CPC | CPM | What you're actually buying |
|---|---|---|---|
| ChatGPT | $3–$5 (recommended max bid) | up to $60 (default max) | Conversational, contextual intent |
| Google Search | $5.26 avg | priced per click | Captured keyword intent |
| Meta | $0.70–$1.92 by objective | $14.19 median | Demand creation at scale |
| $5.58 avg | $33.80 | Verified professional targeting |
On raw cost, ChatGPT's $3–$5 CPC undercuts Google Search's $5.26 average — and Google's CPC is up nearly 13% year over year, with legal clicks running to $8.58 and counting. On CPM, though, ChatGPT's $60 max is expensive next to Meta's $14.19 median. That's the tell: AI inventory is scarce and high-intent, so OpenAI is pricing it like premium real estate, not like social reach.
But here's the contrarian catch, and it's the one number advertisers should stare at hardest: cheaper per click is not the same as better per conversion. A Google Search click converts at a 7.52% average rate because the person typed the exact problem into a search box. A ChatGPT click is cheaper, but the conversion and attribution story is still being written. Early Criteo pilot data claims ~2x conversion rates and ~3x CTRs versus comparable formats — but that's from a tightly curated pilot in three specific categories (consumer electronics, lifestyle, home), with a small set of well-funded brands. Treat it as directional, not as a benchmark.
The honest read for a media buyer: today, ChatGPT Ads is discovery spend, not a performance line item. You test it the way you test a new channel — small, instrumented, learning-led — and you don't pull budget out of Google or Meta until measurement catches up.
This is where the "should they be worried" framing needs a hard dose of realism. Let's put the numbers next to each other.
So the honest scale read is: ChatGPT's ad business is 0.3% the size of Google's, 0.5% of Meta's, and 1.5% of Amazon's. On earnings, this is a rounding error. No one's quarterly numbers move because of a $1 billion ChatGPT ad run rate.
But that's precisely why the moomoo piece's final point — "valuation risk, not earnings risk" — is the right one. Stocks price the future. Google's entire Search franchise rests on the assumption that when people want to research, compare, and buy, they'll keep typing into a box Google monetizes at 89% market share. If even a modest slice of high-value, high-intent queries migrate permanently into AI assistants, investors don't need to see a revenue hit to re-rate the multiple. They just need to believe the trajectory changed.
Ranked by genuine exposure:
Google: the structural risk. ChatGPT overlaps most directly with Google Search on product research, comparison, and purchase decisions — the exact queries that carry Google's fattest CPCs. Google still grew Search 17% in Q2, paid clicks 13%, CPC 3%. There's no earnings cliff. But the ceiling on Search's future is the first thing a skeptical analyst shaves when AI assistants start closing the loop between "ask" and "buy." This is a slow burn, not a crash.
Meta: the budget-shift risk. Meta's Instagram and Facebook content-discovery use cases are genuinely distinct from ChatGPT — you don't open ChatGPT to doomscroll. But Meta lives and dies on performance-marketing budgets, and every performance dollar that moves into AI assistants has to come from somewhere. Meta's exposure is to budget reallocation, not to direct query overlap.
Amazon: the strongest defense. Amazon combines product search and transaction data in one place — it knows what you looked at, what you bought, and what you returned. ChatGPT can influence the "what should I buy" moment, but Amazon still closes the sale, ships it, and owns the purchase record. If anyone can co-exist with (or even use) ChatGPT's top-of-funnel influence, it's Amazon.
Three numbers that settle the debate over the next year:
ChatGPT Ads is the first genuinely new advertising surface in a decade, and the mechanics are more interesting than the headline. OpenAI built a two-door buying system (partners + self-serve), a dead-simple single ad unit rendered beneath the answer, contextual intent matching instead of keywords, and pricing that undercuts Google per click while pricing CPM like premium real estate.
But a $1 billion run rate against $500 billion of incumbent revenue is not a threat — yet. It's a signal. The incumbents aren't about to lose their earnings. They're about to have their multiples questioned, and in public markets, that's often the more expensive problem.
The takeaway for anyone in business, not just in ads: the platforms that capture intent where the decision actually happens win the next decade. ChatGPT just proved it can monetize that moment 200 days after trying. Google's job is to prove the box is still where decisions get made.
