Google Rebuilds the Search Box for the AI Era, the Biggest Change in 25 Years

Google used its I/O 2026 keynote on May 19 to announce what it is calling the biggest change to Search since the search box debuted more than 25 years ago. The reimagined entry point dynamically expands as you type, accepts text, images, files, videos, and entire Chrome tabs as inputs, and is wired into the company's newly general-available Gemini 3.5 Flash model for sub-second responses.

The headline shift is that Search no longer just returns links. For many queries Google now drops users into what it calls AI-powered interactive experiences, including on-the-fly visualizations, mini-apps, and custom UI generated by an agent at the moment a question is asked. The agentic coding piece runs on Google Antigravity, the standalone agent-first development platform that also shipped a 2.0 update at I/O with a new desktop app, CLI, SDK, and managed execution tier — the same stack Search now leans on to spin up bespoke widgets in real time.

The most consequential feature for the long-term shape of Google's business is what the company calls "information agents." These background agents continuously scan the web, financial data, sports scores, and social posts, then notify users when something relevant to a saved query changes. "Provide a synthesized update with links and information you can dive into further," Liz Reid, Google's head of Search, said in the keynote — language that, taken literally, describes an experience in which a user might never need to re-issue the query at all. Information agents launch first for Google AI Pro and Ultra subscribers this summer before broader rollout.

The redesigned box itself starts rolling out this week to all users, with generative UI and mini-app building following in the summer. Pricing for the underlying model is aggressive: Gemini 3.5 Flash now lists at $1.50 input and $9 output per million tokens with a 1M-token context window, materially below what Anthropic and OpenAI charge for comparable frontier models. Google has framed that as a deliberate move to make agentic search economically viable at the scale of its 8.5 billion daily queries.

The competitive read is that Google is choosing to cannibalize its own ten-blue-links business before ChatGPT and Perplexity do it for them. Whether publishers and advertisers will tolerate the trade — fewer outbound clicks in exchange for agents that quietly resolve queries inside Google's surface — is the question that will define the next 12 months for the entire search-economy stack.