One of the most important components of TikTok’s success has been its powerful recommendation algorithm, although its usefulness in the United States is contingent upon a new binding joint venture agreement with ByteDance. Dubbed by some as “TikTok’s crown jewel,” this technology is currently under intense scrutiny due to national security concerns.
In the latter part of 2025, ByteDance signed binding deals to form a joint venture in the United States, headed by Oracle, Silver Lake, and MGX. This deal will transfer control of TikTok’s U.S. app to American and foreign investors, with a planned completion date of January 22, 2026. The aim is to avoid a ban and to separate the handling of U.S. data from ByteDance’s control, while the parent company holds a 19.9% stake.
However, there is still some uncertainty as to the final ownership of the algorithm, considering ByteDance’s previous commitment to wind down TikTok in the United States rather than sell it. As per the agreement, the joint venture will be responsible for the management of U.S. user data, content moderation, and the security of the algorithm, and will also retrain the algorithm exclusively on U.S. data obtained by Oracle. The revenue streams, including advertising and e-commerce, will be handled by a ByteDance subsidiary, with revenue shared with the joint venture.
China’s export control regime in 2020 requires government approval for the transfer of algorithms or source code, making it difficult to share them across borders, and it is unclear what ByteDance’s stance is on this matter. There are also debates about whether ByteDance has completely relinquished control of the technology or simply licensed it, with some comparing Oracle’s role to that of a monitor.
The algorithm of TikTok is characterized by its focus on “interest signals” and not social graphs, a strategy employed by other rival companies such as Meta, which adjusts itself according to the changing interests of users, including their fluctuations on a daily or hourly basis. Along with the short video format and the mobile-first approach, this strategy results in highly personalized feeds, which can give a competitive edge to TikTok over other late entrants like Instagram Reels (2020) and YouTube Shorts (2021).
The complexity of the algorithm is supported by empirical research. A study conducted in the US and Germany among 347 participants, including automated agents, found that the algorithm “exploits” users’ interests in 30-50% of recommendations, showing exploratory content beyond users’ established preferences to improve the algorithm or extend the session length. This serendipitous blending of familiarity and discovery is seen as key to user retention by TikTok executives.
