The Winners of the TikTok Ban

The United States has finally confirmed it will ban TikTok. Three years ago, nobody believed this bill would ever pass — yet it has, and that tells you something significant about how the world's operating logic has shifted.

Douyin's parent company, ByteDance, was founded in 2012 by Zhang Yiming. The company's earliest products were mobile apps for memes and jokes. Building on that early app experience, ByteDance decided to move into news, launching the news app Toutiao. Toutiao's business model and technical capabilities became the foundation for Douyin.

Toutiao succeeded largely because of its strong recommendation engine. Zhang Yiming's key insight was this: search is people looking for information, while recommendation is information finding people. The latter is far more efficient. The simplest approaches to recommendation are content filtering and collaborative filtering. Content filtering is something our brains do every day, though often we still have to consciously sift through things ourselves.

Collaborative filtering, meanwhile, is widely used in the ad-tech world I work in — it's the "you might also like" feature common on social media. The logic is that people similar to you might like the same ad, so we cluster users into groups and generate recommendations accordingly.

Similar ideas around personalization had existed before, but the timing was never right. Back in the era when Yahoo dominated the internet, information online was still fairly centralized. It took the mobile internet revolution and the rise of social media to get users accustomed to a decentralized experience — everyone scrolling through everyone else's individual feeds, growing used to content that felt personal to them. That's the real foundation the recommendation-system era rests on: you need vast amounts of data, users who are used to fragmented, personalized information, and a good delivery vehicle. The mobile era solved all three. Phones are naturally suited to being an information vehicle, and because their screens are small, passive consumption is easier than active input — so we spend enormous amounts of time on them, generating huge volumes of data in the process. That's the fundamental reason a product like Douyin was able to rise.

The Mobile Internet Revolution: Fragmented, Personalized Information

Beyond the basic principles of recommendation filtering, ByteDance's early system drew on three main categories of data: content, users, and context. Each of these could steer recommendations in different directions — users could be segmented in countless ways, content could be judged by whether a user was likely to enjoy it, and context covered things like where someone was and what phone they were using, all of which affected how well recommendations performed. In today's era of large-language-model magic, none of this sounds like black-box wizardry — it feels almost like common sense. But back when ByteDance started in 2012, this was genuinely groundbreaking work.

AI has advanced so quickly that today's recommendation algorithms are far too complex to summarize in a few sentences. But the early recommendation systems were essentially built on the logic described above. Toutiao's success also came down to more than just a strong recommendation engine — the company was also excellent at user acquisition. For instance, Toutiao once spent heavily to get itself preloaded on low-end phones, a strategy that paid off handsomely. ByteDance carried all of this foundational know-how straight into Douyin, simply swapping recommended news articles for recommended videos.

After Toutiao, ByteDance turned to Douyin. Short-video apps already existed by the time Douyin launched, but they were still a fairly niche category — musical.ly, for instance, had built a decent following in North America among trend-conscious users. Douyin wasn't an immediate hit either. But ByteDance kept pouring money into it, and in 2017 the show The Rap of China (中國有嘻哈) gave Douyin its breakout moment, drawing in its first wave of genuinely enthusiastic users. ByteDance also had substantial experience in user acquisition and knew how to grow a user base — and it had deeper pockets than any standalone video-app startup, since it could subsidize growth with cash generated by Toutiao. Eventually it spent $800 million to acquire musical.ly, bringing Asian and American users under one roof for the first time. That's when the Douyin/TikTok we know today truly took shape.

From the user's perspective, Douyin's success rests on highly sophisticated automated content classification. Memes, meanwhile, had already become a normalized subculture — essentially infrastructure for subculture itself — and Douyin emerged as a new channel for meme propagation alongside traditional social networks, tapping into enormous latent user demand.

In 2020, Trump demanded that TikTok be sold, but Biden revoked that executive order in June 2021. Nothing much happened after that until recently — though Trump's team kept a TikTok ban on its long-term agenda. What's different this time is that mainstream voices in the US tech industry are now much more vocally in favor of a ban than they were during Trump's earlier attempt. With that kind of consensus in place, a ban that many expected would only be seriously discussed once a new president took office has instead arrived far ahead of schedule.

YouTube's Video Empire: Heading Toward Near-Monopoly

To figure out who stands to benefit most, it helps to first look at what actually makes Douyin strong. Its real advantage is an extremely high user threshold — by 2024, precision recommendation systems are no longer black magic, so the true moat is an astronomical user base. That means the biggest beneficiaries going forward will be platforms that already have massive numbers of users. By that measure, the clearest winner is YouTube. YouTube was already miles ahead in the video world, and its lead is now likely to harden into something close to an outright monopoly.

Another likely winner is Facebook — its short-video distribution is highly efficient, and the company has never been shy about copying its rivals, having repeatedly out-executed competitors it followed into a market. It's plausible Facebook pulls off another growth spurt. Twitter (X) is a wilder card: it's well suited to fast-moving topic distribution and has a huge user base, but owner Elon Musk's product instincts are unconventional enough that it's hard to predict how the platform will respond. Bottom line: YouTube looks set to become Google's best moat for a good long while, buying Google more time to navigate the upheaval AI is bringing.