Toutiao: Challenging the “One-Feed-Fits-All” Notion of Information Dissemination

A look at how machine learning is changing the way people receive news and other information in China

With the advancement of machine learning, one Beijing-based company, Jinri Toutiao (a.k.a. “Today’s Headlines”), capitalized on the underlying technologies to challenge the “one-feed-fits-all” model of news and information outlets today.[1] Using machine learning to centralize and optimize the distribution of information, Jinri Toutiao (“Toutiao”) created immense value for the modern reader by offering relevant, quality content at a speed and attention level that no human editor is able to achieve.[2]

Launched in 2012, Toutiao became China’s leading mobile app provider of news and other content in just a few short months. Its success is attributable to the machine learning technologies it uses to create newsfeeds specifically tailored to user preferences. According to Vice President Tina Zhao, Toutiao aims “to become the information platform that knows each individual best” with “no two users’ feed lists alike.”[3] Today, Toutiao boasts “~120 million daily active users” who spend, on average, “74 minutes a day” on the app, more than “any other big social platforms in or outside China, including Facebook and WeChat.”[4]

The app generates an initial newsfeed based on information from the user’s profile (e.g. demographics, location, phone model). Then, as users interact with the app, it learns users’ preferences through what they’re reading, time spent on articles, comments and millions of other dimensions. If the user has a tendency to read sports-related news before bed and politics in the morning, the algorithms will learn that nuance and “push” relevant articles to the user during those specific time periods. For most users, the app’s algorithms begin pushing highly relevant recommendations within a day![4]

As with all machine learning, the platform requires massive amounts of data (e.g. user interaction) to provide the level of precision it needs to generate value for users. Thus, Toutiao continuously scours the web for interesting content to enhance user retention and interaction. In 2013, Toutiao introduced Toutiaohao, a self-publishing platform for professional and individual content creators with the goal of boosting its database’s content quantity and quality. The company now claims “1.2 million Toutiaohao accounts, including media outlets, government services, bloggers, and influencers, who publish ~500,000 pieces daily.”[5] Additionally, during the 2016 Olympics, the company introduced Xiaomingbot, an artificial intelligence that published “450 news items throughout the 15-day event.”[6] Although the content was written in simple English, the rate at which the stories went live, often within a few seconds after the end of an event, garnered global attention. Over time, Toutiao hopes to refine the Xiaomingbot technology to produce more compelling stories.

In addition to content quantity, Toutiao must also ensure content quality as it faces the prevalence of fake news and clickbait content. Its parent company, Bytedance, employs a unique battle strategy – fight fake news with more fake news. Bytedance developed two bots, one that writes its own “fake fake news” by learning from the app’s database of real fake news, and another that is “trained to detect fake news by analyzing its counterpart’s fake newsfeed.” According to Toutiao AI Lab head, Ma Weiying, this “dual-learning” process sets the two bots to “…compete with each other” and “push[es] each other to improve.”[1]

While I applaud Toutiao for its track record in innovation and commitment to content quality, the current improvement process is too dependent upon a database of existing fake news, and thus, the solution is always reactive and potentially prone to human manipulation. Going forward, Toutiao should advance its machine learning technology to be able to proactively filter not only fake news, but sensational articles that live within the gray areas. Critics have noticed that there is a disproportionate amount of “soft news” (e.g. celebrity gossip, pop culture, lifestyle articles) distributed through the app. With Toutiao’s reliance on ad sales, its revenue-sharing program with content creators, and the inherent dangers of algorithm bias, the company should implement safeguards to protect the app from developing into an echo chamber of convergent content. Head of Toutiao’s AI Lab, Lei Li, has previously stated that the machine learning model is designed to “expand [users’] horizons,” offering “some content that the user hasn’t signaled a preference for.”[7] While this brings some level of comfort, it is important that Toutiao continuously assesses the integrity of its platform.

Additional Thoughts

With Toutiao as the main source of information for most users in China nowadays, where and how should Toutiao draw the line between a budding social responsibility to balance hard and soft news and delivering its value proposition of providing curated content for users? 

The company recently expanded its censorship team with 2,000 additional hires as Toutiao’s exploding success attracted scrutiny from the Chinese government.[8] Given Toutiao’s immense ability to shape news selection and its colossal database of user data, how should Toutiao navigate China’s complex environment without sacrificing the integrity of its platform?


(793 words)



[1] Hershey, F. (2017). Toutiao’s approach to curbing fake news: teach the AI to write it so that the machines can fight it.. [online] Splice. Available at: [Accessed 10 Nov. 2018].

[2] Hariharan, A. (2017). The Hidden Forces Behind Toutiao: China’s Content King. [online] YC Research. Available at: [Accessed 10 Nov. 2018].

[3] Digipanda Marketing Co. (2018). Toutiao: an Emerging Social Media Platform in China that You Can’t Afford to Ignore. [online] Available at: [Accessed 10 Nov. 2018].

[4] The Economist. (2017). Toutiao, a Chinese news app that’s making headlines. [online] Available at: [Accessed 10 Nov. 2018].

[5] Zhao, X. (2018). Exclusive: Toutiao to build up a partial paywall for its top-ranked news app in China – KrASIA. [online] Available at: [Accessed 10 Nov. 2018].

[6] Cuthbertson, A. (2016). A prolific robot journalist covered 450 Olympic stories. [online] Newsweek. Available at: [Accessed 10 Nov. 2018].

[7] Knight, W. (2017). This Chinese media giant is using machine learning to go after Facebook’s lunch. [online] MIT Technology Review. Available at: [Accessed 10 Nov. 2018].

[8] Feng, E. (2018). Tencent, Toutiao hiring more censors as China increases scrutiny | Financial Times. [online] Available at: [Accessed 10 Nov. 2018].


Clover Food Lab: The Fast-Food Chain that aims to solve climate change via Open Innovation


Printing the Tip of the Spear: Additive Manufacturing at the U.S Military’s Battlefront

Student comments on Toutiao: Challenging the “One-Feed-Fits-All” Notion of Information Dissemination

  1. Thanks for sharing. A lot social media or news app are doing the similar things, recommend news based on past reading record and personal profile, which I find troublesome. The pro of internet is to diverse our views and broaden our perspective however with this so called machine learning, it pushes suggested news to me which indicates my past reading material not necessarily predict my future interests. People also have curiosity to click what’s in front of them. After awhile it’s hard to tell whether we teach machine or AI shows us a pattern to follow.

  2. I agree with you that there is a potential here for Toutiao to use its machine learning technology to filter fake news more proactively. I could see that they would be able to create value in users and society should they be able to help filter fake news on their platforms. With regards to your closing question, I think it would be a very fine balance for Toutiao to take in drawing the line between creating social responsibility and encouraging users to spend more time on the app. It would be very challenging for Toutiao to not include any hard news in its “curated content” to users if some users are only looking for soft news on the app.

  3. I really worry about the question you raised regarding the line between balancing social responsibility and providing people news so curated that it compels them to use JT. With Facebook and Twitter there is already a major concern about the validation of motivated reasoning i.e. the behavior whereby people seek out facts to validate what they already believe in while discounting / disbelieving facts that don’t confirm to their worldview. There is a real danger in our society of people increasingly becoming concentrated in echo chambers that resonate with their belief and social media/news has a real role to play in helping to fix this problem. At the same time, one can argue that the only real role of a business is to maximize profit for their shareholders? At what point does some kind of ethical prerogative become more important than the fundamental job of a firm to be a profit maximizing entity? Should the government be regulating businesses like JT for the public good? Once you go down the slippery slope of “benevolently” inspired social diktats on business, it becomes hard to argue where the line for acceptable government interference really begins or ends. I truly don’t know what the right answer is, but am curious if similar concerns have been sparked in China about JT leading to people becoming less informed and simply more prone to develop hardline views, causing society to become more polarized.

Leave a comment