Top News

Nearly 35% of new websites are AI-generated: Study
28 Apr 2026

A recent study conducted by researchers from Stanford University, Imperial College London, and the Internet Archive has revealed that one-third of all new websites created since 2022 are AI-generated.

The research team published their findings in a paper titled "The Impact of AI-generated Text on the Internet."

The study also found that this influx of AI content is making the web more positive and less verbose.

Research team concerned about AI content's impact on the web
Concerns raised

The researchers were inspired by the 'Dead Internet Theory,' which suggests that much of the internet is now just bots talking to each other.

They were concerned that the rise of AI-generated text could hurt semantic and stylistic diversity, factual accuracy, and other aspects.

Their study found that by mid-2025, about 35% of newly published websites were either AI-generated or AI-assisted.

AI's rapid takeover of the web
Digital transformation

Stanford AI researcher Jonas Dolezal, a co-author of the paper, expressed his surprise at how quickly AI has taken over the web.

He said that after decades of human influence, a significant portion of the internet has been shaped by AI in just three years.

This marks a major transformation in the digital landscape much faster than its initial development process.

Testing critiques of AI-generated text
Research methodology

The researchers also tested six common critiques of AI-generated text, including its impact on viewpoints, disinformation, and citation practices.

They partnered with the Internet Archive to collect website samples from August 2022 to May 2025.

The extracted website text was then analyzed using an AI-detection software called Pangram v3, which had the highest detection rate among several tools tested by the team.

Surprising findings about AI and the internet
Findings revealed

To test if AI was creating an internet full of falsehoods, the team extracted fact-based claims from selected websites and paid human fact-checkers to verify them.

They also computed the outbound link density in AI-generated text to see if it cites its sources.

Surprisingly, only two of their six theories about the effects of AI-generated text were true: it made the internet less semantically diverse and more positive overall.

Future research directions
Future studies

The researchers plan to continue their investigation into how AI-generated text shapes the internet.

Maty Bohacek, a student researcher at Stanford and co-author of the paper, said they are working with the Internet Archive to create a continuous tool that keeps providing this signal going forward.

They also want to look at which kinds of websites are most affected by this trend, broken down by category or language.