Google transforms search into invisible AI assistant working 24/7 across all services

Google stopped being a search engine years ago, but now it's becoming something stranger: an invisible employee living inside your digital life, working around the clock whether your devices are on or off. At its annual developer conference this week, the company unveiled a vision of artificial intelligence so thoroughly woven into its services that the boundary between tool and assistant dissolves entirely.

A year ago, Google announced the end of the internet as billions of people had known it. The company had spent a quarter-century defining how we find information—a dominance so complete that "googling" became a verb. But at last year's I/O conference, executives explained they were replacing the familiar blue links and sponsored results with something closer to a conversation with an AI. This year, they've gone further. The search box itself is being reimagined as a chat window. Users no longer type "Italian restaurants in Madrid" or "flu symptoms." They write paragraphs, ask chains of questions, pose almost philosophical requests. Google calls these searches more "granular and conversational." What they really mean is that the internet is becoming an endless conversation with a machine that sounds confident even when it's wrong.

But search is just the beginning. Google is deploying what it calls intelligent agents—software workers that monitor your interests and concerns continuously, even while you sleep or your phone sits dark on a table. One agent, called Spark, operates in the background like an invisible intern, reviewing bank statements, hunting down forgotten subscriptions, summarizing meetings, drafting documents, and chaining together entire workflows without supervision. Another, Daily Brief, wakes up before you do, scanning your email and calendar to assemble a personalized morning briefing. A third watches financial markets and product availability across multiple stores, alerting you to price drops and restocking. To keep these phantom assistants from feeling entirely unsettling, Google introduced Android Halo—a small visual indicator at the top of your phone screen showing whether Spark is working, waiting, or has a message for you.

The convenience comes with an obvious cost. For these tools to work, you must grant them access to your most intimate digital spaces: every email, every calendar entry, every search, every purchase. Google already knows what you look for, what you watch, where you go, and what you write. Now it's asking permission to read all of it before you do, to anticipate your needs, to manage your anxieties about money and time and forgotten tasks. The company is essentially asking users to trade privacy for automation—to let an AI radiograph their daily life in exchange for having fewer things to worry about.

Google is also reshaping how people shop. Its new Universal Cart monitors prices across different stores and platforms, detecting when items go on sale, when stock returns, and whether products are compatible. It knows your credit cards, your loyalty programs, your spending habits, and the discounts you've forgotten about. The company frames this as finding "hidden savings." It is also, unavoidably, the delivery of more personal data to a corporation that already possesses an extraordinary amount of it.

The transformation extends to YouTube, where a new feature called Ask YouTube turns the platform's search function into a chatbot capable of answering complex questions by weaving together long-form videos, short clips, and timestamps. It's a significant shift for a platform that has already become America's most-watched television service, surpassing cable and traditional streaming. And Google Docs now accepts voice commands, allowing users to compose documents by speaking instructions aloud—a feature that drew sustained applause at the conference.

Underlying all of this are two new AI models. Gemini 3.5 Flash promises to be four times faster than competitors and half the cost, designed to power the agents and automations Google is rolling out. Gemini Omni Flash, meanwhile, generates and edits video from plain-language instructions—ask it to change a background to a science-fiction landscape or add a moving car, and it does so while maintaining visual coherence. It's a demonstration of technical muscle aimed at capturing a share of the digital content wars before rivals do.

Google has also restructured its pricing. The Ultra plan, which cost $250 monthly, now costs $100. A new $20-per-month tier offers access to Gemini 3.5 Flash and YouTube Premium Lite. A $200 option provides access to experimental models and unlimited usage. The company is signaling clearly that it intends to make living inside its AI ecosystem the default—and it's willing to lower prices to make that happen. Most of these features will arrive first in the United States, rolling out to other countries including Spain over the coming months, available only to subscribers. The question that remains unanswered is what happens to the open web—that older internet where you followed links from page to page and ended up somewhere unexpected—when an AI has been assigned to decide what you should actually see.