fg-zettelkasten

don't you know that you're toxic? how influencer‐driven misinformation fuels online toxicity

3 min read

don’t you know that you’re toxic? how influencer‐driven misinformation fuels online toxicity

Summary

This paper investigates whether the source of brand-related misinformation — a social media influencer (SMI) versus an ordinary user — systematically shapes the toxicity of audience responses. Through an empirics-first mixed-method study of 101 Snopes-verified misinformation posts and ~48,821 comments across six platforms (2020–2023), Di Domenico, Mangió, and Dineva show that identical misinformation generates substantially more, and qualitatively different, toxicity when posted by influencers. They argue that influencer credibility and parasocial bonds are not merely persuasive assets but infrastructural mechanisms — operating through legitimation and community enmeshment — that convert audience trust into collective antagonism, typically redirected at the implicated brand. The paper thus reframes online toxicity as a source-amplified, relational phenomenon rather than a property of message content or individual cognition.

Key Contributions

  • Reframes online toxicity around the source of misinformation rather than its content or cognitive susceptibility of receivers.
  • Identifies two influencer-specific mechanisms — brand-related misinformation legitimation (amplifying/sheltering strategies) and community enmeshment (bonding/endearing strategies) — that sustain toxic echo chambers.
  • Develops a typology of five toxicity categories: anti-brand reactions, C2C conflicts, flame-bait firestorms, toxic debunking, and trolling/flaming.
  • Integrates source credibility, parasocial interaction, and social influence theories (compliance, identification, internalization) into a unified framework for influencer-driven harm.
  • Releases a novel multiplatform, cross-industry brand-misinformation dataset and a validated mixed-method protocol.
  • Offers ecosystem-level managerial guidance organized around publishers, platforms, and people.

Methods

A three-stage sequential mixed-method design combining: (1) top-down automated toxicity scoring via Google’s Perspective API (threshold 0.6), extensively validated against Detoxify and Hurtlex; (2) bottom-up Biterm Topic Modeling on the most toxic quartile of comments (n=11,871), producing 41 coherent topics grouped into five categories; and (3) theory-building hybrid thematic analysis of 1,800 representative comments and 45 SMI posts (κ = 0.87–0.96). Logistic regression predicted comment-level toxicity from source type with 15 controls, supplemented by propensity score matching, Gaussian copula endogeneity correction, and category-level OLS regressions with FDR adjustment.

Findings

  • Influencer-posted misinformation elicits ~44% higher odds of toxic comments than identical user-posted content (OR = 0.56 for users; predicted toxicity 3.8% vs. 2.2%).
  • A toxicity–engagement spiral holds for influencers (more engagement → more toxicity) but reverses for regular users.
  • The influencer effect is concentrated in sociopolitical misinformation (OR = 1.87); no significant difference in commercial or health/safety domains.
  • Toxicity peaks on low-pseudonymity, identity-based platforms (e.g., Meta), contradicting the anonymity-fuels-toxicity assumption.
  • Influencer audiences produce homogeneous flame-bait firestorms aimed at the implicated brand; regular-user audiences produce more diverse toxicity (anti-brand reactions, toxic debunking) across multiple targets.
  • Parasocial bonds shield the influencer from blowback by redirecting hostility outward.

Connections

This paper aligns closely with work theorizing the structural and relational dynamics of influencer-driven information disorder, particularly Marwick2025-ov on influencer authority, Frischlich2025-vn on dark participation, and Goel2025-iq on parasocial dynamics in misinformation contexts. Its emphasis on platform-level moderation of toxic spillovers connects to Gonzalez-Bailon2024-rq and Budak2024-ef, while its challenge to cognition-centric accounts of misinformation speaks directly to debates engaged by van-der-Linden2026-jt and Mosleh2024-op. The toxicity-typology contribution also resonates with firestorm and incivility research represented by Rossini2026-jn.

Podcast

A research-radio episode discusses this paper: Listen

Graph View

Backlinks