A comprehensive study analyzing social media content recommendation across Meta, TikTok, and YouTube found that engagement-optimizing algorithms systematically amplify politically polarizing content by factors of 3 to 7 times above its organic reach relative to non-polarizing content. The finding confirms what critics have long alleged and contradicts platform arguments that algorithms are politically neutral optimizers of user interest.
The platforms have disputed the methodology while simultaneously making changes to their algorithmic systems following the publication. Meta has significantly reduced political content recommendations in the wake of public criticism and advertiser concerns. TikTok has introduced more active interest diversification to reduce filter bubble effects. Critics argue these changes are insufficient and that only regulatory intervention requiring algorithmic transparency will produce meaningful behavior change.
