Chatbot ads have moved from experiment to established business for some of the biggest names in AI, and the early results are a mix of clear wins and unresolved risks. Anthropic continues to keep Claude free of advertising, a stance reinforced by its recent IPO filing, which did not raise the prospect of ad revenue. With roughly $112 billion in investor funding, the company can afford that position. The rest of the industry is not waiting.
OpenAI has signed more than 100,000 advertisers in ChatGPT since ads launched there in February. Google recently extended ads in its AI Mode search chatbot to ten additional countries beyond the United States and said users consider the ads highly relevant. Advertisers, however, are not yet enthusiastic about the experience, though they expect it to improve.
Why niche platforms are turning to ads
For smaller chatbot platforms, advertising may be a matter of survival. They lack both the venture capital reserves of leading AI labs and the broader revenue streams of large technology companies. Koah Labs, a San Francisco startup, is building ad technology specifically for AI experiences. The company says it places ads in more than 200 apps, including DeepAI and Liner, and works with brands such as telehealth company Hims and flower delivery service UrbanStems. Koah has attracted about one-tenth as many advertisers as OpenAI, but CEO Nic Baird says the company is learning what works as chatbots increasingly take over tasks once handled by search engines.
AI-written ads draw more clicks
One of the clearest findings from Koah’s network is that ads composed partly by AI generate twice the clicks of those written entirely by humans. The reason is not fully understood, but Baird suggests that AI-generated marketing copy feels more natural alongside AI model output. Koah’s ads are built from product details and images supplied by advertisers, and users do not appear to be seeing hallucinated or incorrect information. Some advertisers, such as drugmakers, face government rules about what they must or cannot say in ads, leading them to keep more human control. According to Baird, those ads perform worse.
The challenge for ad technology companies is to make ads engaging without driving users away or triggering caution among risk-averse brands. That balance may be harder to achieve when humans are removed from the process.
Context beyond the search query
Koah’s targeting goes beyond matching a prompt to a keyword. A Google search for help with stress might surface ads for therapists, but a chatbot can respond with more creativity. Baird says spa ads have performed well within Koah’s network when shown after prompts about stress. The difference, he says, is that Koah can choose and customize an ad based on the prompt, the AI platform’s response, and context from the conversation so far. Chatbot prompts also tend to be at least twice as long as Google queries, giving the company more information for matching ads.
Baird argues that users are used to moving between topics within a chatbot conversation, which may make them more open to adjacent marketing than they are in traditional search. Koah reports click-through rates of about 3 percent, roughly in line with the more mature Google ads system.
Privacy questions remain
When Koah displays an ad, it generally receives the user’s entire conversation up to that point. Traditional search queries usually seek specific information, but people often treat chatbots as conversation partners and confidants. It is unclear whether users will accept the privacy trade-offs that come with personalized chatbot ads as they become more common.
Baird says Koah anonymizes queries and stores them in a de-identified way, but he declined to comment further on data collection and storage. He acknowledges that personalization can become intrusive but argues that there are benefits for users and advertisers as the technology matures. Information from chatbot discussions, for example, could inform personalized content on the page that loads after an ad click. If someone expressed interest in a specific feature or the AI model did not highlight a downside of a competing product, the advertiser could fill that gap. Koah does not have data on whether people want these customized pages, though the tech industry has long argued that people prefer relevant ad experiences. Conversation history could also help advertisers when Koah enables retargeting, showing the same ad to a user again.
Customer support as an ad opportunity
Koah estimates that one in four chatbot queries about a retailer or brand concerns a return, exchange, or other support issue. Baird sees this as a chance for advertisers to address what might politely be called customer success. He expects manufacturers and retailers to eventually target disappointed customers with reminders about warranties, AI-mediated support, and efforts to steer them away from costly customer service interactions. An AI chatbot’s response might provide similar help, but buying an ad guarantees that a company’s message is seen. The best outcome is a faster resolution the user can accept. The worst is a wasted ad buy and a more frustrated consumer.
The agentic shopping unknown
None of the apps where Koah currently serves ads offer AI agents that can shop on their own. But apps such as Instinct and Meta’s Muse are beginning to take over shopping tasks for some users. Meta CEO Mark Zuckerberg recently said Muse could become the company’s largest revenue source, though he suggested that would come from transaction fees rather than ads.
Advertising may still have a role in bot-driven commerce. Early studies suggest ads could influence AI agents. Baird expects an opportunity to give a user’s agent the best possible recommendation, combining a strong discount, a relevant product, and an easy path to purchase. The chatbot must make it abundantly clear to the agent what content is sponsored, he says, or risk losing user trust.
The
