For years, the “dead internet theory” was dismissed as an online conspiracy: the idea that the web had gradually stopped being primarily a place for humans talking to humans and was becoming dominated by bots, automation and synthetic content. In 2026, however, parts of that once-fringe theory are becoming measurable.
The latest research does not prove that the internet is literally “dead” or that most online content is produced by machines. But it does reveal a dramatic shift in the balance between human and machine activity.
A study published in April by researchers including the Oxford Internet Institute examined websites archived by the Internet Archive between 2022 and 2025. By mid-2025, approximately 35% of newly published websites were classified as AI-generated or AI-assisted. That figure had risen from effectively zero before ChatGPT’s launch.
The researchers also found something particularly interesting: increasing AI-generated text was associated with lower semantic diversity and a greater prevalence of positive sentiment. However, they did not find statistically significant evidence that AI-generated text was reducing factual accuracy or stylistic diversity. In other words, the internet may be becoming more synthetic without necessarily becoming uniformly less accurate or less varied.
Newer measurements suggest the transformation is happening even faster. Cloudflare reported that AI-related traffic has surpassed human web traffic, while cybersecurity research has recorded explosive growth in autonomous AI agents interacting with websites. HUMAN Security reported an approximately 7,851% year-on-year increase in agentic AI traffic in its 2026 benchmark.
Meanwhile, a recent Pew analysis of 10,000 webpages found that around 10% showed significant signs of AI authorship in July 2026, compared with roughly 2% five years earlier. When researchers looked specifically at material published after ChatGPT appeared, the proportion rose to around 35%.
Social platforms are beginning to respond. LinkedIn recently introduced a feature allowing users to flag suspected “AI slop”; more than one million people used it within its first two weeks, while LinkedIn said posts identified as AI-generated were subsequently receiving substantially fewer views.
So, is the dead internet theory true?
Not exactly. The evidence does not demonstrate that humans have disappeared from the web. What it demonstrates is arguably more unsettling: humans increasingly share the internet with systems that can generate content, imitate interaction, scrape information and independently navigate websites at enormous scale.
The original theory may have been wrong about the internet suddenly becoming “dead.” But it may have been remarkably prescient about something else: the internet becoming increasingly difficult to tell apart from a conversation being held by machines.
And that distinction could become one of the defining questions of the next decade.