In the past five years since the "pig food theory", LI Liang of ByteDance has refuted the rumor again, and short videos are facing new challenges.

Recently, a rumor about "big tech companies attacking each other" has been widely circulated. It is reported that a vice - president of ByteDance countered the "swill theory" put forward by a vice - president of Tencent, saying, "We're all in the swill - selling business. No one should look down on the other."

Li Liang, the vice - president of the Douyin Group, immediately posted a refutation on Weibo, clearly stating that this was a fabrication and calling on the platform to clean up relevant false information.

After the refutation, netizens' discussions didn't subside. Many netizens said, "Whether it was said or not, I think it makes sense" and "There's really nothing wrong with calling it swill"...

Most industry insiders can easily see through the rumor at a glance. But why do so many netizens believe, forward it, and even think it "expresses their innermost thoughts"? This question might be more worthy of Li Liang's attention than the rumor itself.

01. The spread of the rumor indicates a prior trust issue

Li Liang is the most active executive in Douyin's external communication. From a communication perspective, this rumor hardly meets the basic conditions to "deceive people with false information", but it still spread.

The reason is that it hits a judgment privately agreed upon by a large number of users: In terms of content quality, there is no moral high - ground among major platforms.

The story dates back five years.

In June 2021, at the 9th China Network Audio - Visual Conference, a vice - president of Tencent publicly criticized the short - video ecosystem, leaving the widely - circulated statement: "Personalized distribution is too powerful. If you like swill, all you'll see is swill."

ByteDance immediately sent Li Liang to counter, pointing out that Tencent's WeChat Video was the only short - video platform that hadn't launched the youth protection mode as required at that time, and highlighting the contradiction of Tencent "vigorously developing short - videos while constantly attacking the short - video industry".

This debate had no winner and no conclusion. But its real legacy is the public's pre - set suspicion of the entire industry ecosystem: The moral accusations among platforms are just mutual exposure of each other's shortcomings among competitors, and no one truly stands on the side of users' interests.

Five years later, the reason why this unspoken statement was widely believed is precisely because this suspicion has never dissipated. Moreover, it has sufficient real - world basis.

02. Algorithms not only satisfy preferences but also shape them

The "swill theory" has a key assumption: Some people are naturally fond of low - quality content, and algorithms just satisfy this preference. This logic sounds reasonable, but it ignores a more fundamental fact: Platforms are never passive.

As of December 2024, the number of short - video users in China reached 1.04 billion, with an average daily usage time of 156 minutes, ranking first in the segmented field of network audio - visual applications for six consecutive years. The 156 - minute usage time is not the result of users' "innate nature", but a natural outcome of platforms' algorithms continuously optimizing retention goals and systematically promoting high - stimulus, low - threshold content.

The mechanism is not complex. The business model of short - video platforms is centered around advertising revenue, which depends on usage time, and usage time depends on the content's ability to capture users' attention. Going along this chain, the pursuit of "content quality" by platforms naturally has a certain contradiction with business efficiency. As long as retention is prioritized, the balance of content promotion will continuously tilt towards high - stimulus, low - threshold content. This is not a moral issue of a single platform but the business logic of the entire industry.

The consequences of this tilt have a public - level name. Oxford Dictionary named "Brain rot" as the word of the year 2024, referring to the mental fatigue and attention dulling caused by fragmented content consumption. A meta - analysis covering nearly 100,000 participants in 2025 also found that there is a moderate - intensity statistical correlation (r = - 0.34) between excessive short - video viewing and a decline in cognitive ability, with attention and inhibitory control being the most significantly affected. Although these findings are only statistically related and the causal mechanism is still controversial, the direction is consistent.

The problem is not that "short - videos will definitely make people stupid", but that when the platform's business logic continuously rewards high - stimulus content, the premise of "users' independent choice" has already been quietly rewritten.

This also means that platforms have the ability to break this logic. The 2025 Douyin Financial Content Ecosystem Report shows that the number of financial video submissions increased by more than 55% year - on - year, and the playback volume increased by 50% year - on - year. There is a market for knowledge - based content, and platforms also have the ability to make it visible. As long as the recommendation weight is adjusted, high - quality content can also gain popularity.

The problem has never been the absence of an audience, but whether high - quality content is the default option.

03. The upgrade of forgery: From "poor content quality" to "difficult to distinguish content authenticity"

If the structural problems of platform content still remain at the "quality" level, then the new - type forgery that has emerged in recent years touches on a more fundamental issue: Authenticity.

Staged videos: From "harmless scripts" to consuming social trust

At the beginning of May 2026, a video of less than one minute quickly spread on various platforms. A young girl claiming to be visually impaired was walking on the blind path when an electric bicycle suddenly rushed out and knocked her down. The perpetrator not only didn't apologize but also scolded the girl for "getting in the way". The video quickly triggered an outcry across the network.

On May 16, the Beijing police announced that the video was a false staged production, and the two suspects had been taken into criminal custody in accordance with the law.

After the police announcement, a more complete industrial chain emerged. Mr. Zhang, a visually impaired person, revealed to the media that the visually impaired group he belongs to had received a recruitment notice from an MCN company, clearly requiring "cooperation with the team's shooting plan", "girls are preferred, but boys will also be considered", and stating that "the project can continue permanently if there are no special circumstances". A former employee of an MCN company described it more directly: "This field is saturated. You have to invest in promotion in the early stage and then stage videos... It's just that the MCN company writes the script and then asks the bloggers to play the victim."

There have been similar cases that caused wide - spread on the Internet before, such as the "Qin Lang's exercise book incident".

A commentary in the Beijing News directly pointed out the "upgrade" of this incident: Previous staged videos often created extreme conflicts, while this blind - person - on - the - blind - path incident is more deceptive and harder to identify because it contains real social public issues. The iteration of staged - video technology is systematically raising the identification threshold for platform content review.

AI - generated content: The dual dilemmas of difficult traceability and lagging supervision

If staged - video forgery still relies on real - person appearances and has time - cost constraints, the popularization of AI technology has opened the floodgates for the production of false content at a lower threshold.

In 2025, the scale of China's animated micro - short - drama market (mainly AI - generated comic dramas) reached 18.98 billion yuan, a year - on - year increase of 276.3%. AI face - swapping, AI dubbing, AI screenwriting, and AIGC animated short - dramas have spread rapidly, and problems such as the theft of real - person portraits, vulgar content, and script homogenization have become rampant.

On April 1, 2026, the new regulations on the filing of AI - generated comic dramas by the National Radio and Television Administration were officially implemented, with the core being "filing before launch". One week after the implementation of the new regulations, Hongguo intercepted or removed 3,522 illegal and low - quality comic dramas, and Kuaishou removed about 900 at the same time. There were once rumors in the industry that "tens of thousands of works were removed in a concentrated manner".

The harm of AI - generated rumors is more direct. In March 2025, a 4.5 - magnitude earthquake occurred in Baicheng County, Xinjiang. Only three hours after the earthquake, a rumor that "3 people were killed and 65 houses were damaged to varying degrees" appeared on some platforms, accompanied by AI - spliced pictures of collapsed houses and a "victim's plea for help" video with dialect dubbing. In fact, this earthquake did not cause any casualties. Another false report fabricated by AI, "The mortality rate of those born in the 1980s has exceeded 5.2%", which forged "authoritative data" and "expert interpretations" in video form, triggered collective anxiety on the Internet.

Tencent released an AI detection tool called "Zhuque" at the beginning of 2025, trying to use models to identify false content generated by models. However, regulatory authorities also admitted that "AI review capabilities cannot keep up with the demand for massive UGC content". Facing hundreds of millions of new content items added daily, "review after release" is almost a structural reality in the industry.

In September 2025, the Cyberspace Administration of China and multiple ministries jointly issued the "Measures for Marking Artificially - Generated and Synthesized Content", requiring that AI - generated content must be marked with explicit or implicit watermarks to establish a liability - pursuit chain from the perspective of traceability. This is an important shift in regulatory thinking: From post - event deletion to pre - event marking.

04. Conclusion: From the "quality issue" to the "authenticity issue"

Li Liang's refutation is justifiable and necessary.

However, it is worth noting that this refutation itself has become a real - life case: How false content spreads rapidly on the platform and affects public perception. More intriguingly, the content being refuted is about the content ecosystem of short - video platforms.

From the "swill theory" in 2021 to the exposure of the MCN industrial chain behind the staged blind - person incident in 2026, the governance challenges faced by platforms have changed. In the past, the focus was on whether the content was appealing; now, the more crucial issue is whether the content is authentic.

This is a more difficult problem. The former is more about the battle of aesthetics and taste, while the latter damages the trust foundation of the entire information ecosystem.

The reason why that false statement spread widely is not only because it "sounds reasonable", but also because the public has developed a pre - set distrust of this industry. With more and more staged - video forgeries and the increasing difficulty in distinguishing the authenticity of AI - generated content, users will naturally doubt first and then make judgments.

Once this distrust forms, it is difficult to eliminate it with a single refutation. What platforms really need to repair is not just a single rumor, but the authenticity of the content ecosystem.

Douyin is the undisputed largest short - video platform in China, with a daily active user base of over 700 million. This scale means that every choice ByteDance makes in terms of algorithm logic, content standards, and recommendation weights is not just a business decision of a single company, but actually sets the price for the entire industry: It defines what kind of content "is worth being seen" and what kind of creators "are worth being rewarded".

In the past two years, Douyin has continuously taken actions in content ecosystem governance and made some progress. However, judging from users' feedback after this rumor incident, it is obviously not enough.

The crux of the problem may be that users mainly see platforms' governance actions as "passive responses": refuting rumors when there are incidents, following up when there are policies, and removing content when there is pressure. But the reconstruction of user trust requires active, systematic, and perceptible actions.

For platforms, the crisis never comes from the rumor itself. The public can distinguish between truth and falsehood. The real crisis is that people have already defaulted not to stand on the side of the platform.

This article is from the WeChat official account “Emphasize Next” (ID: leo89203898), author: Yixiu. It is published by 36Kr with authorization.