AI Slop Makes the Internet Fakely Happy

To anyone with a pulse and a smartphone, it’s obvious Internet He has The problem of artificial intelligence regression. The problem has become more serious since the launch of ChatGPT in 2022, with some Social platforms Flooded With AI generated typing. Now, there is data to support anecdotal evidence.

New preprint study Published today Researchers at Imperial College London, Stanford University, and the Internet Archive found that nearly 35 percent of all new websites were either created by AI or with the help of AI. The same study also found that online writing is “becoming increasingly sterile and artificially cheerful.” In other words, AI is making the Internet fakely happy.

The research team tried four different approaches to AI detection before settling on tools from Pangram Labs after they provided the most consistent results. (Although the team found that it performed well in its tests, it’s worth noting that all of the AI ​​detection tools Incomplete.) To collect a representative sample of websites, the Internet Archive was clicked Wayback machinewhich collects screenshots of web pages. In addition to determining the number of sites created between 2022 and 2025 that rely on AI-generated writing, the study also tested six different theories about the characteristics of ramp.

The test on artificial joy examined how artificial intelligence affects writing style online. Using sentiment analysis, which classifies words as positive, neutral, or negative, the researchers found that “the average positive sentiment score for AI-generated or AI-assisted sites was 107 percent higher than for non-AI sites.” Researchers see this rise in artificial happiness as a “symptom” of the “sycophantic and overly optimistic nature of current LLMs”. In this way, the tendency of AI-based writing tools to co-opt their human users has a knock-on effect, making the overall trend of online writing more saccharine.

Another test investigated whether the increase in AI-generated writing reduces the “range of unique ideas and diverse perspectives” on offer. The researchers found that AI has made the Internet less ideologically diverse, with AI sites scoring nearly 33% higher on a “semantic similarity” test than human-created sites.

While these two tests confirmed the researchers’ assumptions about artificial intelligence, the others did not. Four theories tested by researchers have not been confirmed. It is worth noting that they suspected that AI would lead to an increase in misinformation, but their analysis of the evidence did not support the hypothesis. They also guessed that AI writing would not be tied to external sources, and that it would be more general in style than human writing. Confoundingly, none of these theories were supported by evidence either.

While the analysis found that the ideas espoused by writing with AI were more uniform — specifically, more consistently cheerful — writing style It has not been confirmed that it is leveled to the ground. This came as a big surprise to researchers, who had assumed they would see a clear move toward more general production. “Everyone on the team expected this to be true,” says Matej Bohacek, a researcher at Stanford University. “But we don’t have significant evidence for that.”

Before conducting the analysis, the research team conducted a survey on how people feel about artificial intelligence. By comparing the results, I discovered that the researchers were not the only ones whose predictions were reversed. Their study finds that many common beliefs about writing with AI are false.

Like the researchers, most people surveyed also assumed they would experience a rise in fake news as the number of AI-generated websites increases. The vast majority of participants also assumed that writing with AI would cease to be linked to external sources, and that it would have an increasingly common and unified voice. “It’s interesting to see that people tend to expect the worst outcomes,” Bohacek says.

This study is not the final word on what AI is doing to the Internet. “We just wanted to get to work,” says Bohacek, who sees this as a starting point for deeper exploration. As a snapshot of the impact of AI, it provides a particularly human flavor of insight: Sometimes, it’s hard to predict how things will develop.