Koah draws on full chatbot conversations to select ads, WIRED reports

When Koah Labs selects an ad for a chatbot user, it generally receives the entire conversation so far—not just the latest prompt. That practice, detailed in Paresh Dave’s November 5 WIRED reporting, puts private conversational context behind the advertising pitch. One customer, DeepAI, says meaningful ad revenue has not resolved concerns about privacy and design.

The answer becomes an advertising signal

Koah markets its advertising technology as reaching more than 200 apps, including DeepAI’s creative tools and Liner’s citation-focused AI service. Its advertisers include telehealth company Hims and flower-delivery service UrbanStems. The operation offers a view of advertising across smaller AI apps, rather than inside a single chatbot.

CEO Nic Baird told WIRED that Koah can choose and customize advertising using both the user’s prompt and the AI platform’s response. That response carries context from earlier exchanges. He also said chatbot prompts tend to be at least twice as long as Google searches, giving the system more information to work with.

His example is stress: spa advertisements have performed well after stress-related prompts in Koah’s network. Baird argues that people already move between topics within a chatbot conversation, making them more receptive to related offers. That is Koah’s explanation for the opportunity, not a demonstrated measure of users’ comfort with the targeting.

More clicks, with limits on the creative

Koah says users click roughly 3 percent of the ads it shows. Baird also claims ads composed partly by AI generate twice as many clicks as those written entirely by people. Those are company-supplied performance figures, not independently verified results.

The ads draw on product details and imagery supplied by advertisers. But some brands cannot give an AI system broad freedom over the wording. Baird pointed to drugmakers, whose advertisements face government requirements about what they must include or cannot say.

Those advertisers keep more human control, and Baird says their results are weaker. His account describes a constraint on AI-written advertising: the freedom to tailor copy can run into a brand’s legal obligations. It does not establish that regulated advertisers should loosen those controls.

I don’t think the ads are working terribly well for anyone

Kevin Baragona, DeepAI CEO, speaking to WIRED

Revenue has not settled the privacy question

Baragona’s reservations are notable because DeepAI already uses Koah. He told WIRED that advertising revenue has been meaningful, yet he remains uncommitted to the approach. Privacy and design are among the reasons the experience does not feel right to him.

Asked about data handling, Baird said Koah anonymizes the queries it receives and stores them in de-identified form. He declined to provide further details about collection and storage. That leaves an important gap in the public explanation of how the company handles the conversations used to select ads.

Baird sees uses for conversational context beyond choosing the initial advertisement. Nonsensitive details could help personalize the page someone reaches after clicking—for example, by emphasizing a product feature they discussed. Koah does not have data showing whether people want those customized pages.

He also sees a customer-service opportunity. Koah estimates that one in four chatbot queries about a brand or retailer concerns a return, exchange or other support issue. Baird expects businesses eventually to target dissatisfied customers with warranty reminders or AI-mediated support, potentially steering them away from more costly service interactions.