Recently, it can be seen that software and website programming technology has advanced, and with the acceleration of AI, websites can now be developed in a very short time without having to rely on developers.
Recently, through the data collected, it has been announced that 1/3 of the websites developed and recorded since 2022 have been generated entirely by artificial intelligence.
This statement was made by a group of researchers from Imperial College London and the Internet Archive who have used data recorded by the Internet Archive throughout the years 2022 to 2025 to investigate this matter through their paper called “The Impact of AI-Generated Text on the Internet.”.
For this paper, six hypotheses were tested, and all of them show various aspects of AI-generated websites, from the deterioration of ideas and views, writing patterns, content that no longer provides authentic sources of information and others.
The research group took website recording data from the Internet Archive from August 2022 to May 2025. For each URL link sampled, it will be fed into the CDX Wayback Machine API to find the oldest archive recording of the website.
They then used Pangram v3 software to analyze the text on the website to see if it was generated by AI or written by a human writer. They also tested other text analysis software, but found that Pangram produced more accurate and consistent results.
The findings of this paper are quite interesting. Several hypotheses tested were confirmed such as; the deterioration of the views shown on the website, the writing is now more “cheerful” and sterile, and these AI websites no longer show authentic sources for the content shown.
However, one hypothesis that could not be confirmed is related to the Truth Decay Hypothesis, which is that as AI content spreads more and more on the internet, users will more often encounter factually incorrect information and hallucinations.
This group of researchers has only looked for fake news, and interestingly, this seems to be the reason why this hypothesis cannot be tested, as the information displayed on these websites does not pass the standards for it to be defined as sharing fake news.
This hypothesis comes through the Dead Internet Theory which states that human interaction on the internet is “dead”, and most of the traffic generated now is from bot interactions and now, artificial intelligence. One of the biggest examples of this theory can be seen through the social media platform Moltbook which is built specifically for artificial intelligence agents while content posting is completely controlled by search engine algorithms.
In my opinion, this report is quite interesting because it somewhat confirms that the use of AI is now at a level where AI-generated content is not only increasing in recent times, but it is also accepted by the public.
