aifollow.news
Back Rohan Paul
Rohan Paul· @rohanpaul_ai · X· · Original publication time SelectedAI score73

UC Berkeley and University of Washington study estimates AI text on about 24% of crawlable webpagesMachine translation

Automatically verified and published · Generated and evidence-checked automatically; not reviewed by a human.

AI introduction

The study sampled 5.1 million Common Crawl pages from 2019 to 2026 and estimated that about 24% of crawlable webpages contained AI-generated or AI-edited text by August 2026, up from almost none before 2022. A statistical model corrected for detector errors; small, little-linked sites accounted for 78.8% of the rise.

Original post · AI translation

@rohanpaul_ai
Translation · refer to the original

加州大学伯克利分校和华盛顿大学的一项新研究发现,截至2026年8月,约24%的可抓取网页含有AI生成或AI编辑的文本,而在2022年之前,这一比例几乎为零。

研究人员从2019年至2026年的Common Crawl快照中抽取了510万个涵盖各种语言的网页样本。他们以2022年之前几乎不含AI文本的网页为基准,使用统计模型校正了各检测器的误报和漏报。

年度增幅几乎翻了一番:从2023年初至2025年初的4.6个百分点,升至此后的8.1个百分点。规模小、外部链接较少的网站贡献了增幅的78.8%,而网络中最核心的网站最不可能含有AI文本。

Source: Rohan Paul · x.com

Research
Found an error? Send a correction