A recent extensive research project utilizing data from the Internet Archive has revealed that around 33% of websites created post-2022 were produced with the assistance of artificial intelligence, representing a swift evolution in the online landscape. The study, titled “The impact of AI-generated text on the internet,” was a collaborative effort by scholars from Stanford University, Imperial College London, and the Internet Archive. By mid-2025, an estimated 35% of newly established websites were identified as AI-generated or AI-assisted, a significant increase from none prior to the introduction of ChatGPT in late 2022.
The study explored six prevalent theories concerning the influence of AI-generated text on the internet, investigating aspects such as the proliferation of misinformation, the standardization of writing styles, and the accuracy of information presented. Surprisingly, the findings revealed that only two out of the six hypotheses were valid. While AI-generated content was observed to reduce semantic diversity on the web and imbue a more positive tone overall, it did not lead to a noticeable surge in provably false claims nor did it diminish the frequency of citations provided on websites.
The researchers have outlined plans to collaborate with the Internet Archive to transform the study into a real-time monitoring tool rather than a static analysis. They aim to enhance the study’s depth by delving into the specific website categories most impacted by AI-generated text and exploring how these effects vary across different languages.