AI search summaries have transformed how we navigate the web. Google's AI Overviews, Bing's Copilot answers, and Perplexity's approach place synthesized paragraphs atop search results, gradually shifting users from clicking through blue links to accepting pre-packaged answers. This change happened slowly at first, then reached a tipping point.

These summaries flatten multiple sources into single authoritative-sounding text, obscuring nuance and disagreement. Experienced researchers scan result pages because the mix of sources—forum posts versus official documentation—provides crucial context. AI summaries erase this topographical information, presenting confident prose whether the underlying facts are simple unit conversions or complex medical advice.

The verification problem compounds this shift. Users read summaries, find them plausible, and rarely click through to sources. This matters most for medical, legal, and technical queries where wrong answers carry real costs. Websites depend on traffic, and reduced visits weaken incentives to maintain quality content—threatening the web ecosystem itself.

Results-only search puts synthesis back in users' hands. While slower, it produces more accurate outcomes because accuracy requires reading primary sources. AI summaries work for simple factual queries, but they appear indiscriminately on high-stakes searches too. The confidence mismatch creates real risks for critical decisions.