Burying the Lead. At Google I/O, the search giant's annual developers conference, Alphabet-unit Google unveiled its answer to new AI search engines like Perplexity and ChatGPT, which are nipping at its heels. But Google Search's new "AI Mode" was buried beneath a blizzard of other announcements, some more important than others.

Let's sort through it.

Grinding Its Way to AI Relevance. Despite a decade of heavy investment into AI, Google wound up racing from behind when OpenAI's ChatGPT leapfrogged it in November 2022. While we still await some of the most advanced products that were announced at I/O, Google's underlying AI technology seems to be catching up through heavy investment and rapid iteration.

Handicapped. AI Mode is an important step in Google's battle to stave off new competition from AI search engines. It pairs the benefits of AI with the power of Google Search's unparalleled web index. But the implementation shows how the search giant is handicapped in this battle.

Accessing AI Mode requires an extra click, which will likely reduce usage. In designing the new interface element, Google had to be careful not to break the search experience that users know so well -- type and hit return. Perplexity and ChatGPT don't have a legacy interface to serve, and they can make the AI search experience more natural.

Making AI Is Hard, but It's Easier Than Making Products. AI announcements at I/O included AI Mode, Gemini 2.5 Pro Deep Think, Gemini 2.5 Flash, Gemini Live, Gemini in Chrome, Veo 3, Imagen 4, Flow, Beam, Canvas, Agent Mode, and that doesn't include experimental products like Project Aura and Project Astra.

Without looking it up, can anyone tell me what all these things do, and why I should use them? The product branding feels removed from the use case, potentially leading to customer confusion.

This gets back to Google's longtime problem with turning tech into products. As tracked by the website KilledByGoogle.com, Google has shut down 296 products over the years. It has had 13 messaging apps, four of which are still going.

Google's challenge is bringing all these strands together in a way that makes a simple pitch to users, the same way Search did in 1997.

Making Agents Happen. One loud-and-clear message from both Google I/O and Microsoft's Monday Build conference is that AI "agents" are where these companies see the future. Agents are apps that work on top of an AI language model, automating a series of tasks from simple commands. Microsoft is pitching them to its business customers, while Google unveiled Agent Mode for consumer use.

For Google, it's vital to maintain control of these AI personal assistants. Once let loose in a browser, these agents can go everywhere a human goes, including searching on Google. The issue is AI agents don't care about Google's advertising. At some point in the not-too-distant future, a large portion of web traffic may be AI agents. All ad-supported services, including Google Search, should find that troubling.

But if Google is the one providing the agents, it can monetize them, making up for the loss of human traffic on the web.