AI Now Dominates Over a Third of New Web Content, Landmark Study Warns

Breaking: AI-Generated Content Surpasses One-Third of New Websites

More than one-third of all new websites created by May 2025 are now AI-generated or assisted, according to a major study by Stanford University, Imperial College London, and the Internet Archive. The research, published this week, finds that 35.3% of recently published web pages were produced with artificial intelligence—a dramatic shift that brings the long-debated 'dead internet' theory closer to reality.

Of that figure, 17.6% of all new websites were entirely generated by AI with no human editing. The findings, based on analysis of the Internet Archive's Wayback Machine from 2022 to 2025, confirm that the digital landscape is being rapidly reshaped by machines.

“I find the sheer speed of the AI takeover of the web quite staggering,” said Jonáš Doležal, a researcher on the study, in an interview with 404 Media. “After decades of humans shaping it, a significant portion of the internet has become defined by AI in just three years. We're witnessing, in my opinion, a major transformation of the digital landscape in a fraction of the time it took to build in the first place.”

Background: The 'Dead Internet' Theory

The 'dead internet' theory has long been a staple of online conspiracy circles. It posits that the internet, once a human-driven space, has been quietly overtaken by bots and automated content designed to mimic real people. More extreme versions allege that governments and corporations deliberately engineer this shift to manipulate public opinion.

Since the launch of ChatGPT in 2022, the theory has gained mainstream traction. The new study provides the first large-scale quantitative evidence: automated traffic now exceeds human traffic, with Cloudflare and Imperva data showing bots accounted for nearly half of all web activity in 2024.

What This Means: The Internet's AI Makeover

The study tested six common fears about AI-generated content. Only two were confirmed: semantic contraction—a narrowing of diverse viewpoints—and a positivity shift, where online writing becomes uniformly upbeat and sanitized. Both trends threaten the richness of human discourse.

However, the research found no evidence of other predicted harms—no increase in rambling, low-substance text, no single generic writing style, and no widespread lack of citations. Most surprisingly, AI-generated misinformation did not spike in the period studied.

“The absence of some expected negative effects is noteworthy,” Doležal added. “But the speed of change demands vigilance. The internet we knew is being remade before our eyes.”

The study’s authors plan to turn their detection method into a continuous monitoring tool, allowing researchers and the public to track just how far the AI takeover progresses. For now, the numbers suggest a digital future that is increasingly written by machines.