Tag
This arXiv paper analyzes over 5 million webpages from 2019 to 2026, finding that the share of pages with AI-generated text grew from near zero before 2022 to nearly 25% by August 2026. The rise is uneven: English-language and .com domains have higher AI prevalence than other languages or .gov/.edu domains, and AI text concentrates in the least central pages of the web graph, with estimates produced via a Bayesian hierarchical model correcting for detector error rates.
Ten years of DNS measurements show email infrastructure consolidating around two providers, DMARC enforcement plateauing, and a large long tail of infrastructure, raising resilience concerns.