top of page

How Adcash Made ChatGPT Ads Profitable

HomeLibrary

How Adcash Made ChatGPT Ads Profitable

Writer: Niels Moltmaker
Niels Moltmaker
7 minutes ago
4 min read

Ads in ChatGPT are still a relatively new concept. Many advertisers want in, but hesitate to put real budget into such an unproven platform that is still in beta.

At Adcash, we set aside a serious testing budget, and we're now ready to share what we learned. Adcash is an alternative ad network that gives advertisers an option beyond the walled gardens of Meta and Google, particularly in sensitive verticals such as Sports, Gaming and Finance. Our goal with ChatGPT Ads was to generate new advertiser and publisher leads for our marketplace.

The short version: after a rough start, the channel delivered a cost per registration of €6.54, which was 35% lower than Meta Ads. Here's how we got there:

First Impressions: "Kill It Before It Grows"

First impressions matter, and ChatGPT Ads don't make a great one.

The biggest issue is pacing. On day one, our campaigns burned through their daily budget within one to two hours. OpenAI advertises post-payment, and technically that's true. But because budgets are spent so quickly, the first €15 billing threshold is hit almost immediately, and you're charged before lunch.

The €15 threshold matches the minimum daily budget per campaign. That's a steep entry point for smaller advertisers. With three active campaigns, you're committing to at least €45 per day, or roughly €1,350 per month, on a platform that is still in beta.

Controlling Spend: Fewer Levers Than You're Used To

The usual response to a platform overspending is to tighten targeting. In ChatGPT Ads, your options are limited to just four:

  • Excluding devices

  • Excluding regions

  • Editing the ad group's context hints

  • Change objective from conversions to Reach or Impressions (not recommended)

Our analytics showed an unusual pattern. Traffic from ChatGPT Ads had a 90% bounce rate, but the visitors who stayed spent a long time on the page. That points to a split audience. Most clicks ended immediately, while a smaller group engaged deeply. The engaged minority suggests the platform can definitely be used to reach interested readers. The bounce rate, however, suggests the ad placement may be generating accidental clicks.

So we excluded mobile devices on day 2. This quickly solved the overspending problem, as over 80% of ad spend was going to mobile devices, but 75% of conversions came through desktop.

Context Hints Are Fun to Write, But They Don't Move the Needle

ChatGPT Context hints

A feature unique to ChatGPT Ads is context hints: written descriptions that guide which conversations your ads appear in. In theory, this opens up unlimited creative targeting options tailored to your USP.

We experimented extensively. We wrote hints telling the platform to show ads only when competitors were named, and also tried hyper-niche descriptions. We ran several tests to evaluate if the performance changed depending on the prompt. For example, by testing ad groups without a context hint against ad groups with a context hint, or by prompting that the ad should show on certain days and see if there is a lift on those days.

Neither test showed a meaningful difference. In our experience, context hints had little or no measurable effect on performance. Your results may vary, but we wouldn't recommend spending hours perfecting your prompts. The real learning comes from setting the right conversions and the algorithm picking up on the data signals.

A good idea is to start with an “easy” conversion, such as a page view, to give the algorithm time to adjust.

Ad Copy: Don't Get Too Creative

ChatGPT Ad copy

Ad copy, on the other hand, matters a lot. Clear messaging that addresses the user directly and states the offer cleanly worked best. When we got too creative, users couldn't tell what we were offering. In other words, keep it simple.

Our best ads reached a CTR of 1.7%. For context, Similarweb's May 2026 analysis puts the average ChatGPT Ads CTR at 0.68%, with the top quartile of advertisers at 1%.


The image does a different job from the copy. While the copy earns the click, the image is what draws the eye, and gets the copy read in the first place. It needs to stand out. For us, we used a photo of one of our Chief Happiness Officers.

Hitting Target CPA After Two Weeks: The Learning Phase is Real

It was tempting to pull the plug early. But the real strength of ChatGPT Ads only became clear after about two weeks and roughly €500 of spend.

Unlike Google and Meta, ChatGPT Ads has no "learning phase" indicator. You can see it clearly in the performance data, though. After a volatile start, the algorithm stabilizes and begins finding conversions at a strong cost per registration.

ChatGPT Target CPA

Managing ChatGPT Ads feels a bit like watching an AI learn to play Super Mario. At first it runs into walls and gets taken out by every Goomba. Then, gradually, it figures out how to play the game and ultimately outperforms the best humans.

ChatGPT Ads Works Best at the Top of the Funnel

Our funnel analysis shows that ChatGPT Ads works best as a first touchpoint. Of the registrations that involved the channel, 80% had ChatGPT Ads as the first touch and 20% had it as the last touch. Judged on last-click attribution alone, the channel looks weaker than it is.

Looking Ahead: Sponsored Agents

Even in the short time we've had access, the platform has changed significantly. The next significant feature on the roadmap is a feature called Sponsored Agents, currently being tested with select U.S. advertisers.

OpenAI has always stated that ads won't influence ChatGPT's answers. For advertisers, though, having a personal salesperson pitch your product would be a huge advantage. Sponsored Agents seem to be OpenAI's way of squaring that circle. After seeing an ad, users can open aseparate, clearly labeled conversation with the advertiser's agent and ask it questions about the product, while ChatGPT's own answers stay independent.

What to expect when advertising on ChatGPT

  • Control early overspend. Start with a small geographic area, or exclude mobile devices from day one.

  • Don't get discouraged by a rough start. The first days are volatile while the algorithm learns.

  • Budget for the learning phase. Plan to spend at least €500 over two weeks before judging results.

  • Keep ad copy simple and make images pop. Clarity beats cleverness.

  • Judge it on first-touch, not just last-click. ChatGPT Ads works best early in the funnel.


Comments


Subscribe to our newsletter

Featured Articles:

Vector 4.png

About the Author

Niels Moltmaker

Niels Moltmaker is Head of Performance Marketing at Adcash, where he leads performance marketing across the Adcash, Odience, and WYNQ brands.

His approach to marketing is built around a simple principle: great marketers need to see the bigger picture. This means understanding how a product solves real customer problems, connecting marketing efforts across the entire customer journey, and combining creative thinking with data-driven decision-making.

Throughout his career, Niels has applied this approach to scale businesses across technology, B2B, and other industries, turning customer insights and performance data into sustainable growth.

Vector 9.png

Stay connected with us

Join us across our social platforms, where every post, video, and story is crafted to inform, educate and give you expert views of the PPC industry.

Follow us 

Sign up to our newsletter

Stay up to date with PPC & learn from the best in the industry! Subscribe for free to the weekly PPC Live Newsletter.

PPC Live audience seated during a conference session with Connected presentation branding in the background

Early bird ticket sale live! - PPC Live #22 London - Oct 21st - 1:30pm • After party included

bottom of page