How can startups and developers actually monetize their AI products? A startup called Koah, which recently raised $5 million in seed funding, is betting that ads will be a big part of the answer.

If you spend any time online, there’s a good chance you’ve seen plenty of ugly, AI-generated ads — but few to none when interacting with AI chatbots themselves. Koah co-founder and CEO Nic Baird argued that will inevitably change.

“Once these things get outside San Francisco, there’s only one way to make [them profitable] on a global scale,” Baird told TechCrunch over Zoom. “It’s happened time and time again.”

To be clear, Koah isn’t trying to introduce advertising to ChatGPT. (That’s probably something OpenAI will do for itself one day.) Instead, it’s focused on the “long tail” of apps that are built on top of the big models, including apps with a user base outside the United States.

Baird suggested that when consumer AI products were first becoming popular, it made sense for them to focus on “wealthier, prosumer” users, and to monetize those users by converting some of them into paid subscriptions.

But now, someone could build an AI app that reaches millions of users in Latin America, and those users are “not paying 20 dollars a month,” Baird said. So the developer could struggle to bring in subscription revenue, but “they have the same inference costs as everyone else.”

Baird suggested that by successfully figuring out how to make advertising work in AI chats, Koah could actually unlock more potential for “vibe coded” apps that might otherwise be “too expensive to operate at scale” unless their creators raise VC funding. 

In fact, Koah is already serving ads in apps like AI assistant Luzia, parenting app Heal, student research tool Liner, and creative platform DeepAI. Its advertisers include UpWork, General Medicine, and Skillshare.

These ads are marked as sponsored content, and they’re supposed to appear at relevant moments in your chats. For example, if you asking for advice about startup business strategies, the app could show you an ad from UpWork offering to connect you with freelancers who could work with your company.

When Koah talks to publishers, Baird said many of them believe that ads simply don’t work in AI chats, while others have found limited success with AI offerings from older adtech companies like Admob and AppLovin.

But Baird said Koah is 4 to 5 times more effective, delivering clickthrough rates of 7.5%, and with early partners earning $10,000 in their first 30 days on the platform. He added that Koah achieves all that while having less of a detrimental effect on user engagement — though his ultimate goal is for Koah ads to feel relevant enough that they actually improve engagement.

Koah’s seed round was led by Forerunner, with participation from South Park Commons and AppLovin co-founder Andrew Karam. 

Forerunner partner Nicole Johnson echoed many of Baird’s points when discussing the investment over email. She said that when it comes to AI, monetization, is “the elephant in the room amongst builders and investors.” And while the “going standard for monetizing consumer AI services is subscription,” focusing exclusively on subscriptions can “quickly lead to fatigue and churn.”

“Multiple revenue models in Consumer AI are inevitable, and if the past decades of internet services are any indicator, ads will play a major role,” Johnson said. In her view, Koah is “building the essential monetization layer for consumer AI services.”

As for where AI chats fall in the larger advertising ecosystem, Baird and his team have found they represent the middle of the purchase funnel — somewhere between the awareness raising of an Instagram ad and the actual purchase that might be driven by ad in Google search.

“People are not transacting on AI — they’re just not,” Baird said. They might ask a chatbot for recommendations or product details, but then “they’re going to Google to buy.” So part of the challenge for Koah is figuring out the best ways to capture a user’s “commercial intent.”

“It’s not interesting to me to try to figure out, ‘How do we show a display ad in AI?” Baird said. Instead, he wants to understand, “What is the user looking for and how do we give that to them?”