Contents
How Artificial Intelligence is Reshaping the Global Web
Almost a third of all new websites created since 2022 contain content that has been generated or co-created by artificial intelligence. Recent studies are finally shedding light on the true impact this technological shift has had on the digital ecosystem.
While AI has surprisingly made the internet a more “positive” space, it has simultaneously caused a massive decline in linguistic diversity. As the internet is no longer just for humans, with bots and AI generating significantly more traffic, researchers are questioning which of our early assumptions about AI have proven true—and which have fallen flat.
The Rapid Takeover of AI-Generated Content
A collaborative research team from Stanford University and Imperial College London, working alongside the Internet Archive, conducted a massive analysis of web pages. By utilizing the Wayback Machine and advanced AI-detection methodologies, they evaluated a 33-month sample period starting from August 2022.
To accurately gauge the dynamic changes over time, the researchers employed cutting-edge tools like the Pangram v3 detector. This allowed them to classify web content into two distinct categories:
- AI-generated: Content written entirely by large language models.
- AI-assisted: Content co-created or heavily edited using artificial intelligence.
The findings were staggering. By the time of the study’s projections, approximately 35 percent of newly published websites were classified as either AI-generated or AI-assisted. Prior to the public debut of modern generative AI like ChatGPT in late 2022, this percentage was essentially zero.
According to study co-author Jonáš Doležal, the speed at which AI content has “taken over” the web is unprecedented, especially considering this shift occurred over just a few years. This rapid influx has even reignited discussions about the “Dead Internet Theory,” an idea suggesting that the web will eventually be populated mostly by bots interacting with each other rather than human beings.
How AI is Changing the Language of the Web
The research team hypothesized that injecting massive amounts of AI-generated text into the web would fundamentally alter online communication. They looked specifically at semantic diversity, emotional tone, and the average length of digital statements.
Their analysis confirmed several significant shifts in how the internet “speaks”:
- A Decline in Semantic Diversity: Web content is becoming much more homogenous. Articles and posts are increasingly similar to one another, less surprising, and far less varied in how they express complex ideas.
- Emotional “Smoothing”: AI has a distinct tendency to neutralize language. Texts generated by machine learning models are overwhelmingly more positive, excessively polite, and rarely utilize highly negative or confrontational language.
Consequently, the internet is losing much of its natural human friction. The spontaneity, irony, and sharper opinions that have always characterized human-written content are gradually being smoothed out by the safe, corporate-friendly guardrails of popular AI models.
The Truth About Disinformation and the “Death of Truth”
One of the biggest fears surrounding generative AI is that it would trigger an unstoppable flood of lies and digital manipulation. However, the data from this specific study did not confirm that assumption.
While there are legitimate ongoing discussions regarding AI-generated fake news and disinformation being an EU report threat, professional fact-checkers employed for this study analyzed randomly selected factual claims. They found no concrete evidence that AI-generated websites contain more explicitly false information than pages written entirely by humans.
But that doesn’t mean AI has no impact on truth and online quality. The researchers noted the following nuances:
- Unverifiable Claims: AI models frequently generate statements that are practically impossible to verify or fact-check.
- The Dilution of Discourse: There is a massive overproduction of banal, low-information text. While not outright lies, this “fluff” severely dilutes the overall quality of internet discourse.
What Comes Next?
The scientific community doesn’t intend to stop at this single analysis. Researchers are currently planning to transform this localized study into a continuous monitoring tool. This will allow them to track the share of AI content across the global web in real-time, categorized by languages, website types (such as news portals, e-commerce, and blogs), and geographical regions.
Frequently Asked Questions (FAQ)
Does the massive increase in AI content prove the “Dead Internet Theory”?
While the rapid growth of AI-generated content (now present on roughly 35% of new websites) is unprecedented, it doesn’t mean the internet is completely “dead” or devoid of human activity. However, it does highlight a significant shift where machine-generated text is heavily supplementing and sometimes replacing human communication, leading to a more homogenized web.
Why does AI make the internet seem more “positive”?
Most commercially available large language models are programmed with strict safety guardrails designed to prevent offensive, toxic, or highly controversial outputs. As a result, AI tends to write in a polite, accommodating, and overwhelmingly positive tone, effectively “smoothing out” the natural emotional friction and irony found in human writing.
Is AI responsible for a massive increase in factual lies online?
Surprisingly, this specific study did not find evidence that AI-generated websites contain more explicitly false information than human-written sites. However, AI does contribute to the “dilution of discourse” by overproducing banal, low-value content and generating claims that are difficult to verify, which lowers the overall quality of online information.
Source: 404 Media, Heise
Opening photo: Gemini
About Post Author
