“`html
Anthropic has upheld its commitment made during the last Super Bowl: Claude will remain free from advertising. In the company’s recent filing for an initial public offering, there was no mention of plans to sell advertisements to generate revenue from the AI chatbot. Maintaining a principled stance is relatively easy with approximately $112 billion in investor funding at one’s disposal.
However, the chatbot sector may find that selling advertisements is necessary to cover the substantial costs associated with generating AI responses. OpenAI has fully embraced this approach, boasting over 100,000 clients advertising on ChatGPT since it introduced ads in February. Last week, Google expanded advertisements in the “AI Mode” of its search chatbot to ten additional countries beyond the United States, stating that users find them “highly relevant.” Although advertisers are not fully satisfied with their experience, they anticipate improvements.
Advertisements will be crucial for the long-term viability of specialised chatbot platforms, which cannot rely on the same level of generosity from venture capital as major AI laboratories or the vast revenues enjoyed by tech giants. Recently, I spoke with Koah Labs, one of the many emerging companies in the advertising technology space focused on AI-driven experiences. This San Francisco-based startup is promoting its ability to bring ads to over 200 applications, such as DeepAI, which provides tools for creative queries, and Liner, which focuses on delivering quotes within AI responses. Koah collaborates with brands like the telehealth giant Hims and the flower delivery service UrbanStems.
Understanding the Landscape of AI Advertising
While Koah has attracted only a fraction of the advertisers that OpenAI has, its CEO, Nic Baird, believes he is gaining valuable insights into what works and what remains unclear as chatbots replace traditional search engines, and what this surge will mean for user experiences with AI.
Baird claims that ads partially written by AI, as opposed to being entirely human-crafted, generate double the click-through rates. While it is challenging to pinpoint the exact reason, it is likely that AI-generated ad copy appears more natural alongside content generated by an AI model.
Since the ads displayed by Koah are based on product details and images provided by the advertisers, users do not seem to encounter misleading information “hallucinated” by the AI. However, some advertisers, particularly those in pharmaceuticals, face governmental regulations regarding what must or must not be included in their advertisements. Consequently, these companies tend to default to greater human oversight over their ads, resulting in less satisfactory outcomes.
As ads with dull messages or those that seem overly promotional could deter users, advertising technology firms bear the responsibility of striking a balance that captivates users without alarming risk-averse brands. This equilibrium may prove more difficult to achieve without human intervention.
Creative Targeting in Advertising
Searches on Google for “stress relief” are likely to yield ads for therapists. Baird posits that chatbots can be more innovative in their advertising approaches. For instance, spa advertisements have performed well within Koah’s network when displayed following user queries related to “stress.”
The crucial difference is that Koah can select and personalise ads based not only on the query or Google search but also on the context provided by the AI platform’s responses throughout the conversation. Moreover, queries directed at chatbots tend to be at least twice as long as traditional Google searches, giving Koah more context to ascertain which ads are relevant.
Baird notes that users are accustomed to navigating multiple topics within a chatbot conversation, potentially making them more receptive to adjacent marketing compared to traditional searches. Koah reports that users click on approximately 3% of the ads, a figure comparable to the performance of Google’s more established advertising system.
The Role of Context in AI Advertising
When asked to display an advertisement, Koah typically receives the entire conversation up to that point to help determine what to show. While traditional search queries often seek specific information, individuals frequently utilise chatbots not just as conversational partners but also as confidants. It remains uncertain whether users will accept the privacy concessions that come with personalised ads as these become more prevalent.
Baird asserts that Koah anonymises the queries it receives and stores them in a de-identified manner, although he declined to provide further comments on data collection and storage practices. While acknowledging that personalisation can be intrusive, he maintains that as the technology matures, there are benefits for both users and advertisers.
For instance, non-sensitive information gleaned from chatbot conversations could serve as a foundation for delivering personalised content on the page that loads following an ad click. Perhaps someone has shown interest in a specific feature, or the AI model did not highlight a particular drawback of a competing product. The advertiser could then address that gap.
Addressing Customer Concerns through Chatbots
Koah estimates that one in four inquiries to a chatbot concerning a retailer or brand relates to a return, exchange, or other customer service issue. Baird believes that chatbots present advertisers with an opportunity, challenging to replicate elsewhere, to address what could be politely termed a “customer success” issue.
He envisions manufacturers and retailers ultimately reaching out to dissatisfied customers to remind them of warranties, offer AI-mediated customer service, and potentially avert the need for costlier customer service interactions. While an AI chatbot’s response may provide users with similar assistance, purchasing an advertisement ensures that companies convey their message.
The ideal outcome is that users arrive at a resolution more swiftly than they would through other channels. At worst, it results in wasted advertising spend and a more frustrated consumer.
The Future of Advertising in AI-Driven Commerce
Currently, none of the apps where Koah displays ads provide users with AI agents capable of making purchases independently. However, applications like Instinct and Muse from Meta are beginning to take on purchasing tasks for some users. Meta’s CEO, Mark Zuckerberg, recently suggested that Muse could become the company’s largest revenue source, although he indicated that this would stem from transaction commissions rather than advertising.
This does not imply that advertising will vanish entirely from bot-driven e-commerce, as early studies suggest that ads could influence AI agents. Baird anticipates the opportunity to provide the user’s agent with the best possible recommendation, which would consist of a combination of a significant discount, a genuinely relevant product, and a seamless digital purchasing process. “The chatbot must make it ‘absolutely clear’ to the agent which content is sponsored; otherwise, it risks losing the user’s trust,” states Baird.
There is also the possibility that the chatbot advertising market may dissipate before agents take hold if more AI developers adopt Anthropic’s approach. Kevin Baragona, CEO of DeepAI, acknowledges that while advertising revenues have been considerable, he is not committed to them because the experience still seems inadequate in terms of privacy, design, and other factors: “I don’t believe ads are working particularly well for anyone.”
“`