Ads in Chatbots Have Arrived. Here’s What’s Working—and What’s Not

Chatbot ads are no longer hypothetical. OpenAI says more than 100,000 customers have advertised in ChatGPT since February. Google expanded AI Mode ads to ten more countries and claims users find them highly relevant. Anthropic, flush with roughly $112 billion in investor cash and a Super Bowl vow, still omits ads even from its IPO filing.

Niche chatbot apps need a business model that is not pure venture subsidy. Koah Labs, which WIRED profiles, says it brings ads to more than 200 apps and works with brands like Hims and UrbanStems — a smaller advertiser base than OpenAI's, but enough to see patterns.

How this settles will decide whether conversation stays a product or becomes another ad slot with a memory.

We're split between sympathy for the cost problem and wariness of the fix. Generating answers is expensive. Ads are how consumer internet paid for free. Chat is not search, though, and the privacy surface is wider.

Koah CEO Nic Baird's early lessons are concrete. Ads partly written by AI get double the clicks, he says, likely because they blend with model output — while staying grounded in advertiser-supplied product details and images to limit hallucination. Spas after "stress" prompts outperform blunt therapist-keyword matching because the system can use the model's reply and a longer conversational context, not just a two-word query. Koah claims about a 3 percent click rate, roughly Google-like. Roughly one in four brand queries is about returns or support — a lane advertisers want for warranty and deflection messages.

The hard parts are the ones that scale badly. Koah generally receives the entire conversation to pick an ad, then anonymizes and de-identifies storage in ways Baird will not detail further. People use chatbots as confidants. That is a different consent bargain than a search box. Drugmakers and other regulated advertisers keep humans in the loop and see weaker performance. DeepAI's Kevin Baragona says ad revenue has been meaningful and still "doesn't feel right" on privacy and design — "I don't think the ads are working terribly well for anyone."

Agentic shopping complicates it further. Meta's Mark Zuckerberg has pointed Muse revenue at transaction fees more than ads, while early studies suggest ads can still steer agents. Baird wants sponsored recommendations that are obviously marked. Trust dies if they are not.

Anthropic can afford purity for now. Most of the market cannot. Useful free AI may need ads; invasive conversational ads will train users to shut up. I'm watching which side wins the default.

ChatGPT ads launched in February 2026. Google is pushing AI Mode monetization internationally. The rise of shopping agents like Instinct and Muse raises the question of whether the buyer of the future is a human scrolling or a bot being pitched.