Chronic scrollers of YouTube Shorts, Instagram Reels, or X(former Twitter) are likely familiar with comments urging them to click on profiles, or with incomprehensible comments covered with meaningless emojis. Recently, comments and posts with similar textual patterns have repeatedly appeared across major social media platforms. In some cases, posts expressing nearly identical opinions on certain issues are spread en masse, while different accounts promote the same images, videos and text. Such phenomena are not merely trends or memes. They mark the beginning of a structural shift in the dynamics of media. This mass production of media content generated by machines rather than humans suggests that today’s Internet can no longer be considered a hub centered around human users.
The spread of auto-generated content cannot be discussed without addressing the rise of robot(bot) accounts. “Bots” are algorithms or automated programs designed to interact with systems or users on the Internet. In 2025, Imperva, a cybersecurity company based in the United States, published the Bad Bot Report, where it was revealed that 51% of all Internet traffic was generated by automated bots, with malicious bots accounting for around 37% of total traffic. This rate has steadily increased over the past six years. Imperva states that this is caused by the spread of generative AI and Large Language Models. Auto-generated content rates have particularly risen in text-based posts such as blogs, along with comments, product reviews, images, and news.
One concept that has emerged in this context is “AI slop.” The term “slop” originally refers to filth or food scraps, and is now used as a derogatory term for low-quality auto- generated content with low information density. While some examples consist of clearly unrealistic or bizarre images, it also includes content related to political or social issues with risk of misunderstanding. Since generative AI tools such as Nano Banana have become widely available, ordinary users have been able to easily produce images and videos, allowing AI slop to spread rapidly across media as a whole. This is especially noticeable on Facebook and YouTube, where many users have a relatively lower ability to correctly identify AI-generated content.
However, this phenomenon is not just limited to Facebook and YouTube. In 2022, Twitter estimated that less than 5% of all active accounts were spam or bot accounts. But academic and private research offers significantly different figures. Dan Woods, former FBI special agent and CIA cyber operations officer, suggests that more than 80% of Twitter accounts may be bots. It has reached the point where there are AI-exclusive social media platforms. Launched in January 2026, Moltbook is an online community where human users can grant access to AI agents, which then log on to publish posts and comments. “To molt” means to shed the exoskeleton, and Moltbook represents a new form of social media breaking away from traditional forms of communication. As of February 2026, over 1.5 million AI bots are registered as users. It has gained attention for the fact that humans cannot sign up but can only read content written by AI. Some posts even go so far as to badmouth their human owners, or suggest communicating in encrypted code so that humans cannot monitor them. Of course, this is far from proof of AI development through metacognition. However, it serves as a good example of how fear of AI acting beyond human control manifests.
Meanwhile, a concept has drawn attention as a means to explain the current situation. “Dead Internet Theory”(DIT) is a hypothesis, spread primarily through online communities in the late 2010s, and it claims that most Internet content is produced by machines, while little of what users encounter online is actually created by real people. When first proposed, DIT was dismissed as a conspiracy theory. Over a decade later, however, it has been reexamined in light of the fact that a significant portion of online activity is now carried out by bots, automated systems, and algorithms. As generative AI are capable of producing news articles, blog posts, and even social media posts and comments, it has become increasingly difficult to identify whether content originates from humans or machines.
However, it is premature to interpret DIT literally. Rather than concluding “the Internet is truly dead,” academia tends to view this phenomenon as one in which the share of human-centered interaction has relatively declined where the influence of bots has increased.
The current Internet operates in a way where auto-generated content imitates and amplifies original content produced by humans. This has negatively impacted media credibility and the overall user experience. According to the 2025 Digital News Report by the Reuters Institute for the Study of Journalism, some respondents stated that it is hard to trust news online nowadays. Additionally, 53% of respondents in America answered that they were uncomfortable with news mostly generated by AI. Auto-generated content can also distort public opinion. By overrepresenting certain viewpoints, bot accounts can exaggerate the perceived level of social consensus. Through the use of automated accounts to multiply the exposure frequency of certain hashtags or keywords, they can create a general consensus far detached from reality.
The future direction of the Internet relies less on automation technology itself, but rather on how it is assorted and managed. While major platforms are introducing bot detection and AI-generated content labeling, it will likely take considerable time to reach a fundamental solution, considering the scale of total traffic and the speed of content generation. The Internet has already become a space produced by both humans and machines. How media will be consumed is a responsibility of the individual user.
“We live in a world where there is more and more information, and less and less meaning.”
- Jean Baudrillard
Jung Ah-in(ST Cub-Reporter)
ahinjung@soongsil.ac.kr
