ChatGPT Ads early results: CPC, CPM and CTR benchmarks from the US and Europe

Why there’s no “official benchmark” yet

OpenAI told Search Engine Journal plainly in September 2026: it doesn’t yet have performance benchmarks across advertisers, industries or campaign types. The product has changed several times (CPM only, then CPC, then conversions; big accounts, then self-serve) and markets open one by one. Every published number is a snapshot of one account, one sector, one period.

Another trap flagged by analyst Giovanni Perilli: applying a US figure to a European media plan stacks a different currency, market, consent regime and access model on top of already high volatility. So we keep the two strictly apart.

United States: results published by advertisers

Search Engine Journal collected numbers from several US and international advertisers in September 2026:

What these numbers say

The best case is Common Thread Collective: 136,000 weekly impressions, $19,000–$38,000 in attributed revenue for $9,620 spent, a 3.3–6.8x ROAS depending on the attribution model. A solid result for a seven-month-old platform.

The most cautious case is Floyd Blaikie, in B2B: a $9.29 CPC and only 5 organisations matching the ideal customer profile out of 146 identified. The platform reaches broadly; in niche B2B, qualification is still the problem.

Country gaps are striking at Synter: a $5.10 CPC in the UK, $10.62 in the US, $17.59 in Australia and $22.89 in New Zealand. A smaller market doesn’t mean cheaper, especially when few conversations match the targeting.

On CPM, the history speaks: around $60 during the February pilot, CPMs seen near $25 by mid-April, yet still above $60 for some advertisers over the summer.

France: the first published tests

Three French case studies were published within three weeks of launch:

Inside the Studio en-tête test

The most transparent test: launched September 1, 2026 at €25 a day, it logged 19,204 impressions, 398 clicks and €220.04 spent in four days, a 2.11% CTR and a €0.55 average CPC. The clicks-optimised campaign got 334 clicks at €0.51; the conversions-optimised one got 64 clicks at €0.85 and zero conversions. One conversion in total, with an attribution method the agency found unclear.

Lesson: in Europe, start with the Clicks objective. Conversion optimisation doesn’t yet have enough signal to work on a small account.

Gamned! and Havas: melting CPCs

Programmatic agency Gamned!, buying through StackAdapt, reports actual CPMs of €10–12 where announced benchmarks said €15–50, and CPCs of €1.40–€2 against an expected €2–12. Over ten days of optimisation, CPC dropped 50% on average and up to 75% on the best campaign; on a €500 test, CPM fell 33% and CTR rose 20%.

Havas Market, working for French utility EDF on the moving-house topic, describes clicks under €1 and “very competitive” acquisition costs, with a major caveat: very little competition at the time and low volumes of relevant conversations. The agency recommends €10,000–€20,000 for a conclusive test.

ChatGPT Ads vs Meta and LinkedIn

To put these numbers in context, here are 2026 ballpark figures for other platforms (sources below):

Our take: should you jump in now?

Yes, if your business is eligible and your customers ask ChatGPT questions before buying (software, travel, training, equipment, home services…). In newly opened markets like France, the window is favourable: few advertisers, so low bids. It won’t last; in the US, prices rose as competition arrived.

No, if you need immediate conversions on a small budget without a strong landing page: volumes are low and conversion optimisation still lacks signal.

Our advice: a €500–€1,500 test budget, Clicks objective, two or three ad groups built on specific situations, and a landing page that answers exactly the question asked. It’s all in our ChatGPT Ads guide.