Where’s The Business Model in Chat-Based Search?

Two years ago I wrote a series of posts exploring the business model and interface implications of generative AI-based search. At the time, it was not clear how Google would respond to the existential threat that ChatGPT and its peers seemed to present. If it took root, a chat-like interface to search would fundamentally disrupt Google’s core revenue model. What was the company going to do about that?

I noted that six months into the GPT revolution, Google’s response seemed to be overly cautious. I encouraged the famously slow-moving company to go on offense: “It’s time to push something out to market, it’s time to declare yourself the leader in this new market, and it’s time to lay out a vision for what the future of computing will look like,” I wrote. “Imagine if they had waited until they figured out how to make money before launching Google Search?”

Since I wrote that post, Google embarked on what Wired called a “two year frenzy” of “catching up with OpenAI.” The company launched a broadside of AI products, lapped its competitors in most AI-related benchmarks, integrated AI into nearly everything it does, and in the process, doubled its stock price.

But .. has Google really caught up? I’m afraid the answer is “yes…and no.” Which less charitably, could be reduced one more step to…actually, no, but hey, they’ve got good reasons!

Let’s start with the positive. Yes, Google has integrated generative AI capabilities (via its Gemini brand) throughout its ponderously large installed base, from “AI Overviews” in search to “Google AI” in its Workplace suite. It can credibly claim, as it did in its most recent earnings call, that many billions of people now engage with its AI services. Of course, most of those engagements might well consist of folks telling Gemini to stop bothering them (I do this a few times a day), but hey, consumer behavior takes time to shift, right?

It certainly does. But changes to consumer behavior are kind of like bankruptcy. It happens slowly, then all at once. And it’s that possibility that keeps nagging at me as I watch Google attempt to thread the camel of its business model through the eye of an AI needle. The result of those efforts, so far, feels anything but seamless.

Take, for example, Google’s integration of Gemini in its “AI Overviews.” We’ve all noticed the change – by the end of last year, roughly half of Google searches returned an AI-generated synthesis of the familiar blue links (which can still be found below the summary). But while the summaries are often useful, they’re a half measure. First off, as far as I can tell, there’s no paid search (or any kind of sponsored results) inside those AI summaries. And secondly, and most importantly, Google has decided against implementing chat into its main search experience. There’s AI, but there’s not a new experience like one might get with ChatGPT, Perplexity, or Claude.

Claude, for its small part, recently announced that it has incorporated web search into its chat service. This prompted long-time industry observer Om Malik to ask this simple question: Has Search Become Just a Feature? From that post: “Chat just works better than the old way of searching. It’s more human, more intuitive, and more useful. We’re not just witnessing a new feature being added to our digital toolset – we’re watching the emergence of a new way of interacting with information itself.”

Exactly. But Google continues to resist this new approach, because, as I mentioned earlier, the chat interface upends its core revenue model. In one fell swoop, Google could take the lead in chat-driven search* by simply integrating Gemini chat into its main search service, but so far, the company has demurred. Why? Because the company knows that its bottom line, and its stock price, will tank. No one wants to click paid links inside a chat interface.

Over the past two years, Google has run a remarkable sprint and produced commendable new AI products, and is once again considered a leader, if not the leader, in AI. But the ugly truth of its position in the market remains unchanged: It – and the industry overall – still has no clear plan for how to make those innovations pay.

—

*Sure, Gemini is available as a standalone chat app, but as Wired reports, more than 600 million people have downloaded ChatGPT, compared to just 140 million for Gemini’s chat agent. And I’m guessing the engagement numbers for those 140 million Gemini users is likely not very encouraging.

—

You can follow whatever I’m doing next by signing up for my site newsletter here. Thanks for reading.