AI is taking over the web: more than a third of new pages on the internet are created by neural networks.
Analysis of fresh data shows: we are on the brink of a tectonic shift in the structure of the World Wide Web. During a large-scale study covering a random sample of 10,000 English-language web pages collected in July 2026, it was found that approximately 10% of all pages on the internet bear clear traces of generation or significant editing by artificial intelligence. However, this figure is merely the tip of the iceberg.
A New Reality: Synthetic Content Dominates
The key finding that changes the picture is the dynamics. If we exclude from the sample "historical" content created before the era of modern generative models and focus exclusively on materials published after the launch of ChatGPT in November 2022, the share of AI authorship rises sharply. Among such fresh pages, signs of synthetic content are detected in more than a third. This is not just a trend—it is the new norm for the digital landscape, where algorithms have become full-fledged authors alongside humans.
Where Exactly Has AI Settled?
The distribution of artificial content across domain zones is extremely uneven and telling. The commercial sector leads: in the .com zone, signs of AI generation were identified in approximately one in ten pages. This is twice as much as in the non-commercial .org zone, where the figure stopped at 4.6%. The situation with academic and government resources is especially indicative: in the .edu and .gov zones, the share of synthetic content is only about 1%.
Such a disparity is explainable from a practical standpoint. Commercial platforms generate huge volumes of template text—product descriptions, SEO articles, news briefs—which are ideally suited for automation. Meanwhile, the academic and government spheres, with their high requirements for uniqueness and authorship, remain relatively "clean" for now. However, at the early stages of ChatGPT's spread, these differences were less pronounced, indicating an accelerating commercialization of AI content.
Methodology: A Question of Interpretation
It is important to emphasize: the study is not an exact count of pages written by a neural network. The machine learning model used analyzed linguistic markers statistically correlating with AI generation, rather than the document's creation history. Therefore, it is more accurate to speak of an estimate of probable authorship. Additionally, the category includes only fully generated or significantly edited AI text; light editing with a neural network is not taken into account.
My view as an analyst: This data is an alarming signal for the web ecosystem. We are moving toward a state where most "fresh" information on the network will be created without human involvement. This calls into question the reliability of search results and underscores the critical importance of content verification and labeling tools, similar to those recently implemented by Anthropic for its Claude models. Without such measures, we risk drowning in an ocean of plausible but soulless synthetic content.