How Google's AI Shortcoming Became an Advantage

Sometimes, one’s greatest weakness can turn into their most prominent strength. In the field of artificial intelligence, what was once considered a shortcoming for Alphabet (GOOGL)—operating both a search engine and an AI chatbot—has now proven to be a significant advantage. While people do alternate between using search engines and AI chatbots, evidence this year clearly shows that chatbots cannot replace search engines. In many cases, users simply want a quick and straightforward answer—a need better served by search engines than chatbots.

As detailed in today’s coverage, this issue is particularly challenging for OpenAI. The AI leader has invested heavily in developing reasoning models adept at handling complex tasks but overlooked that such models are ill-suited for queries requiring concise, rapid responses. OpenAI’s team has recognized the problem: despite continuous improvements in underlying AI models, user experience has not significantly improved. The reason? Many users turn to ChatGPT not for solving intricate mathematical or scientific problems but for basic inquiries.

For executives at companies like Alphabet, this outcome is hardly surprising. In April, when speculation (including from us) swirled about whether Alphabet would merge its Gemini chatbot with Google Search to avoid user confusion, CEO Sundar Pichai signaled otherwise. During an analyst call, he noted that while there is "some overlap" between search and Gemini, their "use cases are fundamentally different." Similarly, Amazon has faced an analogous situation and likely reached the same conclusion. Its shopping platform maintains a separate search engine distinct from its Rufus AI chatbot. For instance, searching for a specific book is far more efficient using the search engine, whereas Rufus takes longer and may provide excessive, irrelevant information.

OpenAI’s realization that "model improvements don’t always translate to significant user value" highlights an industry-wide risk: companies fixated on advancing AI for its own sake, while neglecting tangible enhancements to product experience, may ultimately veer off course. Today’s report also notes internal divisions at OpenAI between research and product teams, with recent reports suggesting similar tensions at Meta (META). It’s reasonable to assume Alphabet has navigated comparable debates. While clashes between researchers and product teams are nothing new, in AI, the stakes—and consequences—are exponentially higher.