Google has long defined internet search with indexed links, but this year at I O 2025, the company outlined a new future guided by artificial intelligence agents. On stage, company leaders introduced a shift toward an experience where users engage directly with powerful AI assistants instead of traditional search results.
The widespread launch of AI mode brings a conversational assistant to users across the United States, able to browse and summarize web content or guide them in online shopping. For Ultra subscribers, a special AI agent called Project Mariner takes automation further by tackling numerous tasks at once and navigating websites without user input.
AI Agents Transforming the Web Experience
Google’s Deep Research agent, now more personalized and able to tap into Mail and Drive data, digs deep into multiple sites to create robust research reports. Simultaneously, advancements in Project Astra have woven real time visual and audio abilities into both Search and Google Gemini, offering new ways for people to interact with AI.
The move highlights an urgent reworking of Google’s core strategy, accelerated by the disruptive impact of ChatGPT and last year’s rough debut of AI overviews in Search. Early missteps had cast doubt on AI’s readiness, but the latest announcements present a stronger case for AI powered web browsing as the new norm.
Tuesday’s news included the Gemini SDK now supporting Anthropic’s MCP standard, furthering the effort to create a cohesive infrastructure for connecting AI agents to vast online datasets. This change signals Google’s intent to place AI at the heart of how users access information, while rivals like Microsoft promote ideas for an open, agent powered web ecosystem.
However, experts point out challenges that come with the rise of AI agents, especially in the context of the internet’s economic foundation. With agents acting as intermediaries, websites may see fewer human visitors, threatening ad driven revenues for publishers even as commerce platforms seek to benefit.
Additionally, ongoing issues like hallucinations, where AI confidently offers incorrect answers, remain unsolved and were addressed as concerns by Google and DeepMind leaders. Incidents of misinformation not only risk user trust, but also complicate the shift to AI heavy platforms.
Despite obstacles, Google is moving rapidly to reshape the web around AI agents rather than waiting for supporting businesses or technical consistency to catch up. This bold direction marks a significant turning point in both Google’s strategy and the evolution of online discovery.
