Every few years, a new ad platform shows up and the question is always the same: should we move budget? The answer is almost never yes or no. It's usually "test, but know what you're measuring."
ChatGPT Ads hit that moment this month. With over 100,000 advertisers and a $1 billion annualized run rate by August, OpenAI's ad product isn't a beta experiment anymore. It's a line item in someone's media plan. And a recent analysis on Search Engine Journal, citing a working paper from the Frankfurt School of Finance, put some numbers on the table that every PPC team should read carefully.
Carefully being the key word. Because the numbers are real, but the story they tell depends entirely on how you read them.
The headline stat and why it's incomplete
The Kaiser and Schulze paper tracked 102 websites over 60 weeks. The core finding: U.S. paid search generated $2.45 revenue per session, compared to $0.89 for ChatGPT ad sessions. Conversion rates followed the same pattern: 1.17% for search vs. 0.37% for ChatGPT.
When you factor in click prices ($0.72 median for Google Search, $1.72 for ChatGPT from a single agency account), the per-dollar return comes out to roughly $3.40 for Google vs. $0.52 for ChatGPT. The press release called it "less than one-sixth the value for money."
That's a striking gap. But there's a problem with treating it as the final word.
The Google CPC comes from 16 accounts. The ChatGPT CPC comes from one agency (Opascope) that spent about $60,000 over 15 days on campaigns it had already scaled after they showed promise. That's not a median. It's a snapshot of one operator's best-performing work. OpenAI's own suggested starting bid is $3 to $5, which would make the value gap even wider. But it also means the comparison is built on mismatched foundations.
The authors themselves call the work "descriptive rather than causal." That's a detail most of the coverage skipped.
Case studies tell a messier story
Beyond the academic paper, several agencies have shared their own ChatGPT Ads results. And they don't point in one direction.
StubGroup, running B2B compliance campaigns for 30 days, saw a ChatGPT CPC of $0.76 vs. $10.46 on Google. Cost per lead was lower on ChatGPT ($92.21 vs. $157.57). But conversion rates told the opposite story: 0.8% on ChatGPT vs. 6.64% on Google. With a long B2B sales cycle, neither metric alone settles the question.
Workshop Digital spent $2,319 on ChatGPT Ads and got zero qualified leads. Their CPC was $3.14 (less than Google's $6.76 for the same client), but the traffic didn't convert. Brafton spent $1,524, got 263 clicks and one form submission. Their paid media director described ChatGPT Ads as "priced like search, but behaving like paid social."
On the other side, Jotform reported cost per signup roughly 30% below paid search, based on July data. No spend details, no attribution breakdown, but a positive signal from a recognizable SaaS brand.
What you see here is a platform in its first year. The variance is enormous. Some advertisers find cheap clicks that don't convert. Others find pockets that work. That pattern should be familiar to anyone who tested TikTok Search Ads in their early days or ran Google Discovery campaigns before they stabilized.
The bidding system just changed. That matters.
Here's the detail that makes the existing data less predictive than it looks: since November 9, ChatGPT Ads introduced Target CPA and Target CPC bidding. Before that, advertisers were mostly working with Maximize Results (now called Intelligent Bidding), which had no cost target at all.
That's a significant shift. Most of the data in the Frankfurt paper, and all of the case studies, predate this change. It's like judging Google Ads performance based on data from before Smart Bidding existed.
From November 18, legacy API bid values are treated as average-cost targets. Max Bid and Max CPC retire for existing campaigns on January 1, 2027. The bidding infrastructure is being rebuilt in real time. Any performance comparison using pre-November data needs that asterisk.
This doesn't mean ChatGPT Ads will suddenly match Google Search returns. It means we don't know yet what a mature, well-optimized ChatGPT campaign looks like. Nobody does.
What this means for your media plan
At difrnt., we've been running test campaigns on ChatGPT Ads for clients since the spring. The pattern we see mirrors the broader data: lower CPCs than Google for some verticals, but conversion quality that varies wildly by industry and funnel stage.
The comparison that matters isn't ChatGPT vs. Google in aggregate. It's ChatGPT vs. Google for your specific offer, to your specific audience, measured with the same attribution window and the same definition of a conversion. If your Google CPC is $0.72, ChatGPT probably doesn't make sense yet. If your Google CPC is $10, the math changes quickly.
The break-even CPC from the study? About $0.26, with a range of $0.05 to $0.57. That's a useful benchmark: if you can get ChatGPT clicks below that range for your vertical, it's worth running a real test. Above it, you're paying more for less.
In Romania's market, where CPCs have been climbing steadily across competitive verticals, the appeal of a new, cheaper channel is obvious. But cheaper clicks with no conversions aren't cheaper. They're just noise with a receipt.
The search funnel has shifted over the past two years. Users are splitting their research between traditional search, AI assistants, and social platforms. Your ad strategy should reflect that reality. But reflecting it means testing with discipline, not moving budget on a press release.
Allocate 5-10% of your paid search budget to ChatGPT Ads. Run it for at least 30 days. Use the same attribution settings you use for Google. Compare revenue per session, not just CPC. And wait for the new bidding options to mature before drawing permanent conclusions.
The platforms that win long-term are the ones that deliver qualified intent. Right now, Google Search still does that better for most advertisers. ChatGPT might get there. The data doesn't say it won't. It says it hasn't yet.





