For the final 20 years, Google has introduced individuals a listing of algorithmically-selected hyperlinks from the net for any given search question. At I/O 2025, Google made clear that the idea of Search is firmly in its rearview mirror.

On Tuesday, Google CEO Sundar Pichai and his executives offered new methods to deliver customers the net, this time intermediated via a collection of AI brokers.

“We couldn’t be extra enthusiastic about this chapter of Google search the place you possibly can really ask something […] your easiest and hardest questions, your deepest analysis, your customized purchasing wants,” stated Google’s VP of Search, Liz Reid, onstage at I/O. “We imagine AI would be the strongest engine for discovery that the net has ever seen.”

The biggest announcement of I/O was that Google now offers AI mode to every Search user in america. This offers tons of of tens of millions of individuals a button to converse with an AI agent that may go to internet pages, summarize them any method they’d like, and even help them shop. With Mission Mariner, Google is delivering an even more hands-off AI agent to its Ultra subscribers. That agent will deal with 10 completely different duties concurrently, visiting internet pages and clicking round on these pages whereas customers are free to plug away on one thing else altogether.

Google can be making its Deep Research agent, which visits dozens of related web sites and generates thorough analysis studies, extra customized and is connecting it to your Gmail and Drive. In a parallel improvement, the corporate is further integrating Project Astra — the corporate’s multimodal, real-time AI expertise — into Search and Gemini, giving customers extra methods to verbally communicate with an AI agent and let it see what they see.

I may go on, however you get the thought — AI brokers dominated I/O 2025.

The rise of ChatGPT has pressured an AI reckoning at Google, inflicting the corporate to rethink the way it brings customers data from the net. This reckoning actually began finally 12 months’s I/O, when Google introduced AI overviews into Search, a launch that was overshadowed by its embarrassing hallucinations. The rollout of AI overviews made it appear as if AI was not prepared for primetime, and that Search as we all know it was right here to remain.

However at I/O 2025, Google offered a extra compelling, fleshed-out strategy to how AI would reshape Search, and thus, the net. The corporate’s new imaginative and prescient means that the way forward for the net, and the corporate, includes AI brokers fetching data from the net and presenting it to customers in no matter method they’d like.

The concept Google’s AI brokers may change Search is a compelling one, particularly as a result of Google is attempting to put an infrastructure for AI brokers. Google introduced on Tuesday that the SDK for Gemini fashions will now natively support Anthropic’s MCP, an more and more standard customary for connecting brokers to information sources throughout the web.

Google isn’t alone on this shift. At a special tech convention this week, Microsoft CTO Kevin Scott laid out his personal imaginative and prescient for an “open agentic internet,” wherein brokers take actions on customers’ behalf throughout the web. Scott famous {that a} key characteristic to make this attainable can be the plumbing that connects these brokers to one another and information sources — particularly, Google’s Agent2Agent protocol and Anthropic’s MCP.

Regardless of the passion, as Ben Thompson notes in Stratechery, the agentic internet has its issues. As an illustration, Thompson notes that if Google sends AI brokers to web sites as an alternative of individuals, that largely breaks the ad-supported mannequin of the web.

The impacts may differ throughout industries. Brokers will not be an issue for firms that promote items or companies on the web, corresponding to DoorDash or Ticketmaster — actually, these companies are embracing agents as a new platform to succeed in clients. Nevertheless, the identical can’t be stated for publishers, which at the moment are preventing with AI brokers for eyeballs.

Throughout I/O, a Google communications chief advised me that “human consideration is the one really finite useful resource,” and the corporate’s launch of AI brokers goals to provide customers extra of their time again. That will all pan out, however AI summaries of articles appear prone to take {dollars} away from publishers — and doubtlessly devastate the very content material creation on which these AI techniques rely.

Additional, there’s a lingering drawback with AI techniques round hallucinations — their tendency to make stuff up and current it as reality — which turned embarrassingly clear with Google’s launch of AI overviews. Talking onstage Tuesday, DeepMind CEO Demis Hassabis even raised considerations concerning the consistency of AI fashions.

“You’ll be able to simply, inside a couple of minutes, discover some apparent flaws with [AI chatbots] — some highschool math factor that it doesn’t remedy, some primary recreation it will possibly’t play,” stated Hassabis. “It’s not very tough to search out these holes within the system. For me, for one thing to be known as AGI, it will must be way more constant throughout the board.”

The implications might be far-reaching. Widespread hallucinations may lead customers to be extra distrustful of data they encounter on the internet. They might additionally sow misinformation amongst customers. Both final result isn’t perfect.

Google doesn’t appear to be ready for ad-supported companies or AI fashions to catch up — the corporate is pushing forward with AI brokers anyway. Google has probably carried out greater than another firm to steward the net as we all know it. However in what may show a serious turning level, the corporate’s conception of the net appears to be reorienting round AI brokers.


Source link