To anybody with a pulse and a smartphone, it’s apparent that the internet has an AI slop problem. The problem has grown extra extreme since ChatGPT launched in 2022, with some social platforms flooded with AI-generated writing. Now, there’s information to again up the anecdotal proof.
A brand new preprint examine published today from researchers at the Imperial Faculty of London, Stanford College, and the Web Archive discovered that roughly 35 p.c of all new web sites are both AI-generated or AI-assisted. The identical examine additionally discovered that on-line writing is “more and more sanitized and artificially cheerful.” In different phrases, AI is making the web fake-happy.
The analysis crew tried 4 totally different approaches to AI detection before settling on instruments from Pangram Labs after it delivered the most constant outcomes. (Although the crew discovered it carried out properly on its assessments, it is price noting that every one synthetic intelligence detection instruments are imperfect.) To compile a consultant pattern of internet sites, it tapped the Web Archive’s Wayback Machine, which collects snapshots of webpages. As well as to quantifying what number of websites created between 2022 and 2025 lean on AI-generated writing, the examine additionally examined six totally different theories about the traits of slop.
The take a look at that regarded into synthetic cheerfulness examined how AI affected the tone of on-line writing. Utilizing sentiment evaluation, which classifies phrases as constructive, impartial, or detrimental, it discovered that “the common constructive sentiment rating of AI-generated or AI-assisted was 107 p.c increased than that of non-AI web sites.” The researchers see this spike in synthetic happiness as a “symptom” of the “sycophantic and overoptimistic nature of present LLMs.” On this approach, AI writing instruments’ tendency to suck up to their human customers has a spillover impact, making the general tenor of on-line writing extra saccharine.
One other take a look at investigated whether or not the enhance in AI-generated writing shrinks “the vary of distinctive concepts and various viewpoints” on provide. The researchers discovered that AI did make the web much less ideologically various, with AI web sites scoring roughly 33 p.c increased on testing for “semantic similarity” than human-made web sites.
Whereas these two assessments validated the researchers’ assumptions about AI, others did not. 4 theories that the researchers examined had been not confirmed. Notably, that they had suspected that AI would lead to an increase in misinformation, however their evaluation of the proof did not help the speculation. That they had additionally guessed that AI writing wouldn’t hyperlink out to external sources, and that it might be stylistically extra generic than human writing. Confounding expectations, neither of these theories had been supported by the proof, both.
Whereas the evaluation discovered that the concepts espoused by AI writing had been extra homogenous—and particularly, extra constantly cheery—the writing model itself was not confirmed to be flattened. This got here as a giant shock to the researchers, who had assumed they’d see a transparent transfer in direction of extra generic output. “Everybody on the crew anticipated that to be true,” says Stanford researcher Maty Bohacek. “However we simply don’t have important proof for that.”
Prior to conducting its evaluation, the analysis crew commissioned a ballot on how folks really feel about AI. Evaluating it to the outcomes, it found that the researchers weren’t the solely ones who had their expectations upended. Many generally held beliefs about AI writing are mistaken, their examine finds.
Like the researchers, most individuals polled had additionally assumed that they’d encounter an increase in pretend information as the quantity of AI-generated web sites they noticed elevated. The overwhelming majority of respondents had additionally assumed that AI writing would cease linking to external sources, and that it might have an more and more generic, uniform voice. “It’s attention-grabbing to see that folks tended to count on the worst outcomes,” Bohacek says.
This examine is far from the final phrase on what AI is doing to the web. “We simply wished to break floor,” says Bohacek, who sees this as a jumping-off level for deeper exploration. As a snapshot of AI slop’s affect, it presents a very human taste of perception: Typically, it’s merely exhausting to predict how issues will unfold.
Disclaimer: This article is sourced from external platforms. OverBeta has not independently verified the information. Readers are advised to verify details before relying on them.
