ChatGPT Ads Hit $1B in 200 Days. One Agency's Own Test Got 58 Clicks and Zero Leads.

OpenAI said on August 31 that ChatGPT Ads reached a $1 billion annualized revenue run rate in under 200 days, and opened self-serve buying across 31 European markets. Roughly $83 million a month, from a product that started testing in February.

Set that beside something we published on July 13, when eMarketer capped the entire chatbot advertising market near $5 billion against OpenAI’s own $100 billion pitch to investors. One platform now claims a fifth of what analysts sized the whole category at.

Both numbers can be right at once, and the space between them is where your test budget lives.

What OpenAI announced

The milestone came with a geographic expansion rather than a performance disclosure. Self-serve access went live across 31 European markets. Ads now run in more than 40 countries. India, one of ChatGPT’s largest markets by weekly active users, launched last week with 50 brands and agency partnerships with WPP and Omnicom.

“Reaching $1 billion in ARR in under 200 days shows the scale of the opportunity ahead,” said Dave Dugan, OpenAI’s VP of Global Ad Solutions.

Dugan framed the European move around access: “Expanding self-serve access across European markets opens that opportunity to businesses of every size, from startups building their first campaign to global brands looking for new ways to grow.”

Read what’s in the announcement and what isn’t. Revenue, countries, brand counts, timeline. No click-through rates, no conversion rates, no cost per acquisition. That’s normal for a platform at this stage, and it’s also the reason a run rate on its own tells you nothing about whether to spend.

The number OpenAI didn’t publish

Choice OMG, a Canadian agency, ran its own ChatGPT campaign in June and published the results. Managing Director Peter Jaffray’s team reported 8,940 impressions and 58 clicks, a 0.65% click-through rate, at average costs of CA$4.42 and CA$4.88 per click on the two CPC-bid campaigns.

Then the line that matters:

“Lead forms submitted from those 58 clicks: none.”

One test, small volume, one vertical, and a single campaign proves nothing on its own. Jaffray’s own writeup cites a larger SE Ranking test that saw more than 97,000 impressions and 1,263 clicks at a 1.30% click-through rate, double his. The honest read of both is that reach is real and demonstrable while conversion is unproven in public.

That’s the gap the $1 billion doesn’t address. A run rate measures what advertisers spent, not what they got.

Why spend is running ahead of proof

Three things pushed money in fast, and none of them is evidence the ads work.

The barrier fell. The $50,000 minimum from the spring launch is gone. Early CPMs around $60 have come down toward $25 in some cases, with recommended starting bids of $3 to $5 per click. A test that cost a serious commitment in March costs a few hundred dollars now.

The tooling filled in. Since the May self-serve launch OpenAI has added a conversion-optimized objective, custom audiences from email and phone lists, and geographic targeting down to the ZIP code. Each addition makes the platform legible to a media buyer who already knows Meta and Google.

And the holding companies moved. WPP and Omnicom signing on for the India launch does more than add spend. It puts ChatGPT on the consideration list in planning meetings where nobody has run a test yet.

What the analysts are watching

eMarketer’s Nate Elliott, its principal AI analyst, put the constraint on the ad-tech side rather than the demand side:

“OpenAI can eventually get to parity [with Google and Meta] in terms of adtech partnerships if they want. But it’ll take time—and given the company’s burn rate, impending IPO, and astoundingly aggressive ad sales targets, time seems to be the one thing they don’t have.”

That pressure is worth understanding, because it shapes what the platform does next. A company that needs advertising revenue quickly has an incentive to loosen targeting, widen inventory and raise ad load. We watched OpenAI remove the daily budget cap earlier this year, and take ads into Europe with personalization off by default to fit the consent rules there. Both moves changed what a campaign could do without changing what OpenAI reported.

What this means for your budget

If you have not tested, the case for a small test is better than it was in March, mostly because it got cheap. Treat it as buying information, not buying customers.

Three rules make that test worth running.

Set a floor for the answer, not the spend. 58 clicks cannot tell you whether a channel converts. Decide up front how many clicks you need before you’ll believe either result, then fund that number or skip the test. Most teams fund a dollar amount and get an unreadable outcome.

Instrument it yourself. Platform-reported conversions are the weakest evidence available anywhere, and on a platform with this much revenue pressure they’re weaker still. Tag the campaign, watch it land in your own analytics, and compare against a channel you already trust.

Judge it against your real alternative. The comparison that matters is ChatGPT against the next thousand dollars of whatever already works for you, not ChatGPT against doing nothing.

The $1 billion is a genuine milestone and a real signal of advertiser interest. It is not yet a signal of advertiser results, and OpenAI has not published anything that would settle that. Until somebody does, the strongest public data point is still an agency that spent real money, got 58 clicks, and wrote down that none of them filled in a form.