Charles Zhang’s Following Feed: Who Pays for His Internet Ideal?

In recent years, Charles Zhang has been promoting the “following feed” on Sohu Video. His idea is to reduce the influence of algorithmic recommendations, let users follow people who interest them, build relationships through genuine follows, reposts and interactions, and take some of those online relationships into the offline world. This approach speaks to people’s frustration with excessive recommendations, competition for attention and superficial social interaction. It has an appealing idealism. Whether that ideal is viable, however, ultimately depends on whether it can sustain itself within the internet as it actually operates.

To begin with, a following feed is hardly a new way of distributing content. Early social networks grew out of users choosing whom to follow, subscribing to updates and reposting content. Platforms later expanded algorithmic recommendations not simply because their operators preferred algorithms, but because users were following more people and more content, making selection necessary within limited screen space and attention. Even if Sohu leaves users to decide whom to follow, the platform still has to make choices about content ranking, public discovery channels, trending topics and how new users discover content. The difference between a following feed and existing recommendation platforms is chiefly the balance between users choosing what to receive and platforms deciding what to distribute, rather than a fundamental distinction between having algorithms and having none.

More importantly, the commercial conditions driving this shift have not disappeared. A platform that depends mainly on advertising faces continuing pressure to expand its user base, increase usage frequency and keep people on the platform for longer. Algorithmic recommendations have become more powerful precisely because they work well towards those ends. Reducing recommendations might earn Sohu some users’ goodwill, but it could also weaken its ability to generate advertising revenue. If it gradually strengthens recommendations in pursuit of profit, it may retrace the path other platforms have already taken. Unless Sohu finds a different revenue structure, simply reducing the proportion of recommended content does not amount to a new business model.

Zhang also stresses meeting online and getting together offline. Yet following someone online and socialising offline often serve different purposes in a relationship. Before people know each other well, public information helps them observe and understand one another. Once they become friends, or develop a closer relationship, their conversations may move to private channels. People who meet through Sohu can easily switch to WeChat afterwards. Offline relationships may, of course, generate content that flows back onto the platform, but that return is not inevitable.

The following feed may therefore resemble a transit point for relationships, with people continually arriving and leaving, rather than a reservoir that steadily retains user activity. The more successfully the platform helps people form offline relationships, the more pressing one question becomes: once those relationships exist, why should people keep interacting publicly on Sohu over the long term? If the answer is to keep organising offline events and earn revenue through sponsorships and service fees, the online platform gradually becomes a customer acquisition channel for an offline business. Its commercial centre of gravity has shifted.

The following feed also offers only limited barriers to competition. Other platforms can readily copy following, reposting and social interaction based on shared interests. Sohu may be able to cultivate distinctive communities, but if users can easily move their relationships elsewhere and community activity depends on a constant supply of newcomers, the following-feed mechanism alone is unlikely to provide a strong competitive moat.

Sohu began exploring related social products around 2018, while the specific form of the following feed emerged gradually later on. Including those early attempts, the effort has lasted close to eight years. When an internet project has been explored for so long, yet its immature commercialisation is still explained chiefly as a matter of “cultivating the market”, it should at least confront a practical question: does the market still need educating, or is the demand itself too limited to support the commercial scale being sought? Some people liking a social platform with fewer recommendations and a stronger emphasis on relationships does not mean enough people will pay for it, or that the platform can earn enough revenue to cover its costs.

Sohu’s financial position makes the question sharper. According to its publicly released financial results for 2025, the group reported annual revenue of approximately US$584 million, of which online games contributed about US$506 million, or roughly 87%. The Changyou gaming segment reported operating profit of about US$237 million, while the Sohu business segment recorded an operating loss of around US$331 million. The group as a whole still recorded an operating loss of approximately US$94 million. The Sohu business segment includes multiple businesses and expenses, however, and separate revenue, cost and profit figures have not been disclosed for the following feed. The entire segment’s losses therefore cannot be attributed to the following feed, nor can these figures establish how much it has actually lost.

What these figures do reveal is Sohu’s economic structure: games are a major source of group revenue and operating profit, and gaming profits offset some of the losses from media and other businesses in the consolidated operating results. Zhang’s ability to persist with a social initiative that has yet to demonstrate independent profitability is closely connected to the group’s gaming business and existing financial resources. A project’s continued existence does not mean the market has validated it commercially. Sometimes it simply means the parent company can afford its costs.

There is also a contrast here between the ideal and the accounts. Zhang criticises the internet’s excessive pursuit of traffic and hopes to restore more natural human relationships. Yet a major source of the profit supporting Sohu’s operations is a gaming business that also competes for users’ time and spending. Games are not inherently immoral, of course, and earning money from them does not by itself prove that Zhang’s idealism is insincere. The question worth examining is this: when a vision that claims to offer a different internet ecosystem cannot, over an extended period, establish a financially self-sustaining model, and instead depends on profits from the existing ecosystem, is it a model others can adopt, or a personal ideal subsidised by other businesses?

Entrepreneurs are certainly entitled to use profits to support causes they believe in, even to run unprofitable projects for years. But presenting such a project as an alternative path for the internet requires more than a vision, relationships and idealism. It also requires answers about where the revenue comes from, who bears the costs, why users stay, and why this path will not lead back to stronger recommendations and fiercer competition for traffic.

Eight years of exploration cannot logically prove that the following feed will never succeed. It is, however, long enough to demand more convincing evidence than a promise to “keep cultivating the market”. The market has no obligation to fulfil an entrepreneur’s ideals, and a founder’s conviction does not make a business model viable. What Zhang needs to demonstrate is not the depth of his dissatisfaction with today’s internet, but whether his proposed alternative can stand on its own without continuing subsidies from another business.