Recently, "Slope" was included as a hot word in Webster's Dictionary for 2025, defined as "low-quality digital content typically generated in bulk by artificial intelligence (AI)". According to experts, "Slope" meant "soft mud" in the 18th century, "food residue" in the 19th century, and later extended to "garbage" and "almost worthless products". In Chinese, it is translated as "swill" and "dirty water". Experts believe that "Slope" has become an annual buzzword and an authoritative dictionary specifically defines "AI garbage", accurately projecting the public's discomfort and dissatisfaction with the massive amount of low-quality AI content, just like seeing sewage, making people instinctively want to retreat. After the rise of artificial intelligence big language models, people quickly noticed that big language models often fabricate seemingly real and reasonable information, such as citing data and materials from seemingly authoritative research reports and academic papers for analysis and argumentation. However, these reports, papers, data, and materials do not exist and are all fabricated by AI. This phenomenon of big language models' serious nonsense 'is called the' AI illusion '. Compared to the "AI illusion", "AI swill" can be described as "serious nonsense", with a large amount of low-quality content generated by AI flooding the internet, causing homogenization, blurred authenticity, massive information accumulation, and even falling into a vicious cycle of "swill feeding swill, garbage generating garbage". Its severity and harm should not be underestimated. How is' AI swill 'generated? As the exploration of this issue continues to deepen, platform companies and regulatory authorities have discovered a hidden black and gray industry chain - promoting products by creating junk information and feeding it to AI, and even using fictional products, fake evaluations, and industry analysis to make AI recommend non-existent brands. These types of services generally do not directly "advertise" AI, but instead systematically lay out content, repeatedly "feed" information, and affect AI citation paths. These false or low-quality content feeds are one of the sources of harmful junk information generated by AI, and also a key focus of "AI waste" governance. To curb and crack down on such black and gray industries, it is urgent to seize key clues and take precise and effective punitive measures. In recent years, as an important component of digital reading and national reading, online literature reading in China has achieved explosive growth. The "Report on Integrated Development of Publishing (2025)" jointly compiled by 10 publishing think tanks, including the China Academy of Press and Publication, shows that in 2024, the revenue scale of Chinese online literature will reach 49.55 billion yuan, and the cumulative user scale will reach 638 million. In the process of rapid development, the problems existing in some online reading platforms have become increasingly prominent. The platforms continue to improve their governance capabilities, comprehensively prevent and resist the intrusion of junk information, and become another key point in curbing "AI garbage". In the AI environment, readers' aesthetic appreciation for high-quality content on online literature platforms becomes even more precious, and their demand for high-quality content is becoming increasingly strong. However, contrary to this, low-quality AI content has become a "bait" for some platforms to attract users and expand traffic, not only occupying readers' limited reading time and paying the price of fragmented attention, but also shaking the foundation of the "content centric" industry. In particular, some free platforms use AI for mass production of low quality and highly homogeneous web articles, intentionally or unintentionally acquiesce in the use of AI for foam creation, resulting in a flood of "AI swill", which has caused serious damage to the digital original soil, digital content ecology and online text reading environment. The "Regulations on the Promotion of National Reading" proposes that "the state encourages the development and application of new technologies, new carriers, and new facilities to promote national reading, supports the use of information technology to carry out national reading promotion activities, and expands the influence of national reading promotion activities." As a new technology, new carrier, and new facility for national reading, online cultural platforms should adhere to the principle of "helping readers grow" and promote national reading with high-quality development of content ecology. They must be vigilant about "traffic first", avoid excessive dependence on AI and algorithms, avoid relying solely on flow quantification to drive out bad money and good money, and avoid becoming a "AI swill" disaster area. (New Society)