Behind the screen: The use of Facebook accounts with inauthentic behavior during European elections

Summary

This paper examines the use of inauthentic Facebook accounts to amplify political messages on the official pages of Romania’s four leading parties and alliances (PSD, PNL, AUR, USR) during the 2024 European Parliament election campaign. Working around Meta’s restrictive API policies through manual data collection, the authors apply Oprea’s Authenticity Matrix to a corpus of 4,476 shares of the most-viral posts from each page. They find that hyperactive users (HAUs)—accounts sharing the same post four or more times—account for a substantial share of distribution (up to 45% on individual posts), and that these accounts display recurring profile patterns. The paper argues that this constitutes a form of public-opinion manipulation that violates Meta’s own Community Standards yet goes largely unaddressed during elections, exposing a gap between platform policy and enforcement.

Key Contributions

• Empirical documentation of HAU-driven amplification on official Romanian party pages during an EU-level election.
• A descriptive profile of dominant HAU characteristics (no disclosed affiliation, exclusively political content, missing human profile features, very few friends, etc.) usable by researchers, fact-checkers, and ordinary users.
• Demonstrates that the Authenticity Matrix can serve as a viable manual detection instrument in API-restricted environments.
• Provides evidence supporting calls for share-count limits, expanded researcher data access, and stronger co-regulatory enforcement under EU instruments (DSA, Code of Practice on Disinformation).

Methods

Between December 2024 and May 2025, the authors manually collected share data from the three most-shared posts on each of the four party/alliance pages during the May 10–June 8, 2024 campaign window, yielding 4,476 viewable shares (6.7% of 70,293 total). Users were classified as Normal, Moderately Active, Hyperactive (≥4 shares), or Super-active (≥10 shares). Each HAU was then scored on Oprea’s Authenticity Matrix—a 0–100 Likert instrument across personal info, account activity, and likes/interactions—with cross-coder checks. Analysis is qualitative-comparative, supplemented by descriptive statistics.

Findings

• HAUs accounted for 23.2% of all analyzed shares; SAUs reached 33.9% on one USR post.
• Per-page HAU ranges: PSD 0–17.1%, PNL 0–26.2%, AUR 20.6–28.1%, USR 14.2–45%—suggesting all major parties benefited from inauthentic amplification, but to differing degrees.
• Extreme behavior observed: single accounts sharing the same post 22–31 times; posting up to 187 items/day or 75 posts in 10 minutes.
• HAU profile traits: 95.5% hid political affiliation; 61.6% showed bursty/inactive cycles; 40.2% posted only political/civic content; 18.8% had no human profile/cover photo; 17% had <10 friends; 7.1% used non-human names.
• Authenticity Matrix classification of HAUs: 38.4% authentic, 42.9% borderline, 18.8% clearly inauthentic.
• Results are consistent with prior HAU proportions reported in earlier European political-engagement studies.

Connections

The study contributes country-specific evidence to the growing literature on coordinated and inauthentic amplification during elections, complementing platform-level and cross-national analyses such as Kulichkina2026-zk, Schroeder2026-im, and Minici2024-tf. Its feature-based, manual detection approach contrasts with the network- and ML-driven detection pipelines explored in Luceri2025-tr, Goel2025-iq, and Mannocci2025-ig, while its concern with the gap between platform policy and enforcement resonates with Graham2025-gp and Graham2026-fb. The Romanian case also extends the geographic scope of CIB research beyond the more frequently studied Anglophone and large-platform contexts addressed in works like Yang2025-iv and FitzGerald2025-nv.