35% of the Web is AI: Stanford Research and ALT Effect
Tech 35% of the Web is AI: Stanford Research and ALT Effect The AI wave shaping the future of the internet is materializing with new data: By mid-2025, 35% of newly published websites carry AI-generated or supported content. Research by Stanford University, Imperial College London, and Internet Archive reveals that this rate was nearly zero before ChatGPT‘s November 2022 launch. Researchers examined 33 months of site snapshots from Internet Archive’s Wayback Machine and classified each page using the AI text detector Pangram v3. This rapid rise, as put by Jonáš Doležal from Imperial College London, proves that the web, shaped by human hands for decades, has fallen under AI dominance in just three years. Stanford Research Method: 33-Month Data Analysis Researchers scanned millions of pages from web archives to calculate the AI ratio. Pangram v3s 95% accuracy in classification makes the findings reliable. This method separates AI texts by lexical and semantic features; for example, repetitive patterns and low entropy carry AI traces. AI Content Hypothesis Tests: Semantic Similarity and Sentiment Score Researchers tested six hypotheses about AI content on the web with data; only two were confirmed. AI sites scored 33% higher in pairwise semantic similarity scores compared to human-written ones, meaning the same ideas