TikTok Algorithm Prioritized Pro-Republican Content in Three States Before 2024 Election

A study published Wednesday in the journal Nature found that TikTok's recommendation algorithm systematically prioritized pro-Republican content on the For You pages of dummy accounts set up to mimic users in New York, Texas and Georgia in the period leading to the 2024 U.S. elections.

Researchers created 323 dummy accounts and conditioned them to behave like real users by having the accounts watch videos aligned with either the Democratic or Republican parties. They then monitored the videos recommended on the accounts' For You pages, the platform's primary feed, for 27 weeks of the presidential campaign.

More than 280,000 recommended videos were reviewed using a combination of human and AI analysis. The study reported a consistent imbalance in the content shown. Accounts trained on pro-Republican content viewed about 11.5 percent more content agreeing with their views than pro-Democrat accounts did.

Accounts trained on pro-Democratic content were about 7.5 percent more likely to be shown pro-Republican content than the reverse.

The researchers stated that Democratic accounts were shown significantly more anti-Democratic content than Republican accounts were shown anti-Republican content. Pro-Democrat accounts received disproportionately more cross-partisan videos on immigration and crime.

Pro-Republican accounts saw more cross-partisan content focused on abortion. The accounts were located in the three states through mock GPS and virtual private network routing. New York was selected as a strongly Democratic state, Texas as a strongly Republican state and Georgia as a battleground.

The researchers cautioned that the findings should not be generalized beyond these three states.

The study noted that TikTok's For You page is almost entirely driven by the platform's algorithm based on behavioral signals such as watch time. Users do not need to follow any accounts for the system to generate recommendations. This setup minimizes user self-selection compared with the main interfaces of other social media platforms, according to the paper.

The authors acknowledged that the study examined the types of political content users are exposed to but did not analyze the influence of the videos on political beliefs or behavior. The bots captured only the early stages of a user's experience on the platform and analyzed English-language video transcripts, which would not capture cues conveyed through visuals or other languages.

The researchers said studying the extent to which political content can be skewed on such feeds remains relevant to debates about platform transparency and algorithmic accountability. The study pointed to the European Union Digital Services Act, which requires large online platforms to assess and mitigate risks to electoral processes.

About 42 percent of U.S. social media users say these platforms are important for getting involved with political and social issues, according to a Pew Research finding cited in the paper. The study noted that the demographic of users ages 18 to 29 shifted 10 percentage points toward Trump between the 2020 and 2024 elections.