Google DeepMind CEO Demis Hassabis expressed surprise at OpenAI’s decision to introduce ads into its AI chatbot, suggesting the move may be premature and could damage user trust. Speaking to Axios at the World Economic Forum in Davos, Hassabis said his team at Google has been weighing the issue “very carefully” and is not feeling pressure from Google’s core advertising business to rush into the same path. His remarks come shortly after OpenAI announced it would begin testing ads as a way to monetize the portion of ChatGPT’s 800 million weekly active users who do not pay for subscriptions.
Hassabis said he was “a little bit surprised” OpenAI has moved so quickly toward ads, even though he acknowledged that advertising has historically funded much of the consumer internet. “If done well, they can be useful,” he said, but he also warned that the dynamics change when the product in question is an AI assistant. “If you think of the chatbot as an assistant that’s meant to be helpful…there is a question about how ads fit into that model,” he said, noting that users need to trust their assistant. The implication is that any commercial interruption could undermine that trust, especially as AI tools become more personalized and integrated into daily life.
Hassabis reiterated that Google currently has “no plans” to add ads to its AI chatbot, though he said the company would continue to monitor how the market responds to OpenAI’s move. He also highlighted the difference between search and chat-based AI. Google Search, he argued, is built on an established understanding of user intent, which makes ad placement easier to justify. Chatbots, by contrast, are meant to evolve into personalized assistants that know users deeply and can help across multiple areas of life, making the insertion of ads more sensitive and potentially intrusive.
The discussion comes amid a growing wave of consumer backlash to ads in AI. OpenAI previously tested a feature that suggested apps during conversations, and users reacted negatively, saying the suggestions felt like intrusive advertising. OpenAI later turned off the feature, claiming it had “no financial component.” Hassabis’ comments suggest that even if money is not changing hands, the mere presence of promotional interruptions can degrade the user experience and erode trust.
At the same time, Google is advancing its own AI personalization features. The company recently launched new capabilities in AI Mode that allow users to opt in to having Gemini tap into Gmail and Photos for tailored responses, similar to Gemini’s Personal Intelligence feature that draws from Gmail, Photos, Search, and YouTube history. The move underscores Google’s strategy of improving usefulness through personalization, but it also raises questions about how data-driven relevance will coexist with the company’s advertising model.
Hassabis acknowledged that there may be a way to integrate ads into AI products “right” at some point, but he stressed that Google is not under pressure to make hasty decisions. “We don’t feel any immediate pressure to make knee-jerk decisions like that,” he said, describing DeepMind’s approach as scientific and rigorous. The comment reflects a broader tension in the AI industry: the need to monetize increasingly expensive infrastructure without undermining the trust and utility that have made these systems popular in the first place.
