After all, AI models are a lot like human beings.
The new study from the University of Texas at Austin, Texas A&M, and Purdue University shows that large language models fed a diet of popular but low-quality social media content experience a kind of “brain rot” This may sound familiar to those who have spent too much of their time scrolling on X and TikTok.
“We live in an age where information grows faster than attention spans—and much of it is engineered to capture clicks, not convey truth or depth,” Junyuan Hong says, the incoming assistant Professor of the National University of Singapore worked on the research as a UT Austin graduate student. “We wondered: What happens when AIs are trained on the same stuff?”
Hong and colleagues pre-trained two large open language models with different types of texts. The models were then fed with a mixture of text that was highly readable and not. “engaging,” Posts on social media that are widely shared and contain hyped-up or sensationalized text, such as “wow,” “look,” You can also find out more about “today only.”
Researchers then measured the impact using several benchmarks. “junk” Social media diet: Meta Llama & Alibaba Qwen
The models fed junk text experienced a kind of AI brain rot—with cognitive decline including reduced reasoning abilities and degraded memory. According to two measurements, the models became more psychopathic and less aligned with ethics.
These results are consistent with research conducted on humans. shows Low-quality content online has been a problem for many years. detrimental effect On people’s cognition. Pervasiveness and the prevalence of this phenomenon were seen “brain rot” Oxford Dictionary word of the year in 2024.
Hong says that these results are critical for the AI sector, since model-builders may believe social media postings are an excellent source of data to train their models. “Training on viral or attention-grabbing content may look like scaling up data,” “He says” “But it can quietly corrode reasoning, ethics, and long-context attention.”
It is especially alarming that LLMs are suffering from brain decay, when AI itself is increasingly creating social media content. Much of it is optimized to engage. Researchers found that low-quality material had a negative impact on models, and they could not be easily improved by retraining.
The results also indicate that AI systems developed around social platforms like Grok could suffer from issues with quality control if the user-generated content is used to train the system without a consideration for the integrity and authenticity of the post.
“As more AI-generated slop spreads across social media, it contaminates the very data future models will learn from,” Hong says “Our findings show that once this kind of ‘brain rot’ sets in, later clean training can’t fully undo it.”
This edition includes Will Knight’s AI Lab newsletter. Read previous newsletters here.

