Google’s AI Search Is Getting Smarter—But Is It Killing the Open Web?

When Google CEO Sundar Pichai sat down for a wide-ranging interview on the Decoder podcast, the conversation was designed to explore the company's reimagining of Search for the AI era. Instead, a live demonstration produced a moment of rare candor that has since become emblematic of a much deeper, unresolved conflict between the world’s largest search engine and the publishers who depend on it for traffic.

During the interview, host Nilay Patel ran a live search for “best Chromebook.” The results were revealing. Google’s AI Overview confidently declared a single device—the Acer Chromebook Plus Spin 714—as the winner, while lower down the page the organic results, including a thread from Reddit, offered a different recommendation . Confronted with the discrepancy in real time, Pichai acknowledged the flaw: “I think it’s probably more opinionated than it should be for the particular query” . He speculated the result might have been personalized to Patel’s search history, but the admission was clear: the AI system had been too authoritative on a subject that demanded nuance and a range of options .

This wasn’t just a product glitch. It was a window into the structural tension at the heart of AI-powered search.

The 38% that publishers can’t ignore

Pichai’s admission landed in the middle of a heated industry debate over whether AI-generated summaries are helping or hurting the web. The most damning evidence came from researchers Saharsh Agarwal and Ananya Sen, who published a working paper in April 2026 presenting the first randomized field experiment on the topic. Their methodology was a marked improvement over prior correlational studies: with a treatment group exposed to AI Overviews and a control group that wasn’t, the researchers could isolate causation. The finding was stark—AI Overviews caused a 38% drop in organic clicks on queries where they appeared . This was presented as causal proof of a trend publishers had been measuring observationally for over a year.

Google’s framing of the data couldn’t be more different. In the same Decoder interview, Pichai argued that as its technology improves, lower-quality clicks are being filtered out, describing the shift as “a natural evolution” . Separately, Google’s VP of Search Liz Reid has repeatedly described AI Overviews as pruning “bounce clicks”—visits where users hop onto a page for a quick fact and leave immediately. Reid insists these are precisely the visits AI Overviews eliminate, preserving deeper, more meaningful traffic .

There is a crucial gap in this explanation: Google has not shared publisher-facing data to verify that only shallow visits are being lost. Pichai’s own admission complicates Google’s narrative even further. If an AI Overview is confidently delivering a single “best Chromebook” answer that differs from organic results, then the traffic it’s replacing isn’t necessarily low-value bounces—it could be editorial traffic from trusted reviews and comparisons.

I/O 2026: Moving from a directory to an agent

If the click debate was already intense, Google’s announcements at I/O 2026 turned it into something existential. Google presented what it called the largest overhaul of Search in more than 25 years, unveiling an agentic transformation that goes far beyond AI summaries .

The company upgraded Search to use Gemini 3.5 Flash as the default model for AI Overviews and AI Mode, bringing frontier-level performance for coding and agentic reasoning to billions of queries . More importantly, the vision for Search shifted from a directory of links to a system that can think, act, and complete tasks on a user’s behalf. Google described the change as moving into “the era of Search agents, where you can easily create, customize and manage multiple AI agents for your many tasks, right in Search” . Instead of sending users to a site to book a service, the agent might book it. Instead of linking to a spreadsheet tool, Search might generate one on the fly.

This transformation pushes Search decisively away from the classic “10 blue links” toward AI-powered interaction . Combined with the existing evidence on click loss, it intensifies rather than resolves the core conflict. The field experiment showed AI Overviews reduce outbound clicks even in their current, comparatively static form. If the new agentic Search experience answers questions and completes tasks without requiring a click, the trend line for publisher traffic can only steepen.

The bottom line

Sundar Pichai’s “more opinionated than it should be” remark may have seemed like a small product note, but it illuminates the entire problem. On subjective queries, AI-generated search results are already overstating a single answer. Meanwhile, a randomized experiment shows a 38% drop in organic clicks, and Google’s roadmap points toward a future where the search engine acts more as an autonomous agent than a bridge to the web. The company’s stance—that declining clicks simply represent quality filtering—fails to account for what happens when the answer is wrong, or when users stop looking beyond it at all.