AI Overviews Risk Killing the Open Web

Łukasz Grochal

A recent Pew Research Center report examines Google’s AI Overviews feature—AI‑generated summaries displayed at the top of search results—and finds it significantly alters user behavior. In March 2025, 58% of surveyed U.S. adults encountered at least one AI Overview. On pages with these summaries, users clicked traditional search links only 8% of the time, compared to 15% without AI Overviews. Just 1% clicked links cited in the AI summaries.

Users were also more likely to end their search session immediately—26% did so versus 16% for regular results. Although these AI summaries often reference reputable sources like Wikipedia, Reddit, YouTube, or government sites, most users do not visit those sources. The practice is raising concerns among publishers and open‑web advocates, who argue it undermines web traffic and content discovery. The share of searches generating AI Overviews increases with query complexity: appearing in only 8% of one‑ or two‑word searches, but in 53% of ten‑word queries or questions starting with "who," "what," "when," or "why."

References
3 sources
01
searchenginejournal.comSearch Engine Journal
02
9to5google.com9To5Google
03
windowscentral.comWindows Central
Qwen3.6 Coding Agent Benchmarks Chart Visual

Exploring Qwen3.6: Coding Benchmarks and Speed

Palantier Dilemma Human Rights vs Sercurity

Europe's Palantir Boom Amid Sovereignty and Rights Fears

Project Glasswing: Anthropic Mythos Zero-Day Exploit Finder Art

Claude Mythos Leak Ignites Fears of Unstoppable AI Exploits

OpenRouter LLM Leaderboard April

Chinese AI Models Dominate OpenRouter Top Six in Token Usage

Claude Code’s Big npm Leak

Inside the Claude Code Leak and Anthropic’s Agent Design

China AI accelerator card shipments vs NVIDIA 2025 chart

NVIDIA’s AI Chip Share in China Drops from 95% to 55%

TurboQuant KV Cache Compression Visualization

Google’s TurboQuant makes AI caches smaller and faster

Black Forest Labs FLUX.2 klein

FLUX.2 klein 9B-KV Explained: Speed, Quality, GPUs

Nvidia Slashes LLM Context Memory With KVTC Design

KVTC: Nvidia’s 20x LLM Memory Cut Without Retraining

OpenAI Sora shutdown concept

Sora’s Short Life: Inside OpenAI’s Quiet Retreat