The diffusion and reach of (mis)information on Facebook during the U.s. 2020 election

Summary

This paper provides one of the largest-scale empirical characterizations of how content—and misinformation specifically—diffused on Facebook during the U.S. 2020 presidential election. Drawing on diffusion trees from roughly one billion re-shared posts by U.S. adult users between July 2020 and February 2021, the authors compare misinformation against other content categories in terms of virality, audience reach, and the structural role of specific actors in propagation. They argue that while most posts reach limited audiences, misinformation is disproportionately represented among the small set of posts that achieve viral, mass-scale diffusion, and that identifiable platform actors and structural features play an outsized amplifying role.

Key Contributions

• One of the largest empirical studies of platform-scale information diffusion during a major U.S. election, leveraging privileged data access to Facebook re-share cascades.
• A comparative framework that distinguishes misinformation diffusion from the diffusion of other content categories, rather than studying misinformation in isolation.
• Empirical evidence informing the long-running debate about whether misinformation is uniquely viral, and about which actors and structures concentrate exposure to it.

Methods

The authors construct diffusion trees from approximately 1 billion Facebook posts re-shared at least once by U.S.-based adult users over a seven-month window covering the election and its aftermath. Posts are classified into misinformation versus other content categories, and propagation dynamics—virality, cascade depth/breadth, and audience reach—are measured comparatively across cascades.

Findings

• The distribution of reach is highly skewed: most posts reach few users, while a small subset goes viral at mass scale.
• Misinformation exhibits distinctive diffusion dynamics relative to other content types, particularly in how reach accumulates across cascades.
• Viral misinformation exposure is concentrated, with a limited set of platform actors and sharing patterns disproportionately responsible for amplification.

Connections

This work anchors a growing strand of platform-data studies of misinformation diffusion and concentrated amplification; it pairs especially closely with DeVerna2025-dl and Bollenbacher2026-vz on cross-platform diffusion dynamics, and with Mosleh2024-op and Budak2024-ef on the concentrated role of specific actors and the prevalence question. It also speaks to broader debates about the actual scale of misinformation exposure addressed in Renault2025-uh and Graham2025-gp.