Problematic Health Information Online
The Shape of the Problem: From Exposure to Ecosystem
A first cluster of papers under this topic pushes back against the assumption that low-quality health content saturates the average user’s feed. Lyons2026-ca combines survey data with passive web and YouTube tracking to show that exposure to low-credibility health content in the US is, in aggregate, rare — yet startlingly concentrated: the top 10% of users account for 77% of exposure, and older adults dominate that tail. Crucially, referrals come not from search or social platforms but from other low-credibility sites, suggesting habituated cross-domain consumption rather than algorithmic accident. This reframes the policy problem away from “everyone is drowning in misinformation” toward a model of niche but consequential audiences moving through parallel information environments.
Scalco2026-bd supplies a complementary, supply-side lens. Rather than measure what users see, it asks when the credible doesn’t show up: it operationalizes “information voids” as statistical anomalies in the balance between demand (searches, Wikipedia views) and supply (Facebook, Twitter, news) across six European countries during the COVID-19 vaccine rollout. Voids persist longer than overabundance episodes and correlate with sharp drops in highly credible content and rises in misinformation. Read alongside Lyons2026-ca, the picture sharpens: dedicated consumers seek out low-credibility ecosystems, and moments of credible undersupply open windows in which problematic content fills the gap.
Coordination, Superspreaders, and Asymmetric Networks
A second strand foregrounds the actors and structures that move content. Efstratiou2026-ij traces COVID-19 preprints across Twitter and news media and finds a coordinated retweet network that is 96.4% contrarian, amplifying a small set of credentialed anti-consensus experts who would not otherwise rank near the top of the conversation. Bot scores don’t distinguish these coordinated accounts — the manipulation is human and strategic. News coverage tends to follow superspreader activity, with high-trust outlets aligning with conformist superspreaders and low-trust outlets with contrarians: a bifurcated information pathway from platform to press.
Song2025-yh extends the coordination story to Facebook, applying CooRnet-style coordinated link sharing behavior (CLSB) analysis to UK and US vaccine discourse. Both pro- and anti-vaccine camps coordinate, but the political-cultural framing of their messaging diverges sharply: UK anti-vaccine CLSB concentrates on safety and trial skepticism (a frame attached to the AstraZeneca controversies that Scalco2026-bd flags as void triggers), while US anti-vaccine CLSB centers on individual freedom and religious exemption. The paper’s normative move — that coordination is ideologically agnostic and also undergirds credible NHS/CDC messaging — usefully complicates the “coordinated = inauthentic” reflex that has dominated platform policy. Together with Efstratiou2026-ij, it suggests that the asymmetry lies less in whether actors coordinate than in what kind of source material they pull into circulation.
From Speech to Outcomes
Bollenbacher2026-vz is the most direct attempt to close the loop between online content and offline harm. Using a SIRVA compartmental model fit to county-level data on cases, vaccinations, and geolocated antivaccine tweets, the authors estimate that antivaccine Twitter exposure caused roughly 14,000 vaccine refusals, ~545 cases, and ~8 deaths in the US over six months of 2021 — a lower bound, but causally identified via shuffling tests and Bayesian model comparison. Read against Lyons2026-ca’s findings of concentrated exposure, the estimate may seem modest; but it materializes the mechanism that voids (Scalco2026-bd) and contrarian amplification networks (Efstratiou2026-ij, Song2025-yh) are otherwise theorized to produce.
Movements, Mainstreaming, and the Wider Discursive Drift
Rothut2026-or zooms out from health proper to ask what protest mobilization around health policy does to the broader information ecosystem. Tracking the German Querdenken movement on Telegram over two years, it shows how anti-COVID-measures protest channels became structurally embedded in far-right and conspiracist networks: outlinks shifted from mainstream toward alternative/partisan outlets, and a community of far-right brokers — not Querdenken itself — occupied the network’s center. The argument for “protest-facilitated mainstreaming” articulates a mechanism the other papers gesture at: shared anti-elite framing, more than ideological commitment, is the connective tissue that lets contrarian health communities (Efstratiou2026-ij), liberty-coded anti-vaccine coalitions (Song2025-yh), and dedicated consumers of cross-domain low-credibility content (Lyons2026-ca) be drawn into adjacency with extremist milieus.
An Emerging Synthesis
Taken together, these papers trace a layered account of problematic health information online. At the user level, exposure is concentrated rather than diffuse (Lyons2026-ca). At the ecosystem level, asymmetric supply–demand dynamics create temporal openings for misinformation (Scalco2026-bd). At the network level, coordinated communities — frequently anchored by credentialed contrarians — exploit those openings and shape downstream news coverage (Efstratiou2026-ij, Song2025-yh). At the behavioral level, the cumulative effect is measurable in vaccination rates and epidemic outcomes (Bollenbacher2026-vz). And at the political level, the same communicative infrastructures serve as conduits for the mainstreaming of radical frames well beyond the original health topic (Rothut2026-or). The open methodological frontier — visible across the set — is integrating these scales: linking void detection to coordinated amplification, coordinated amplification to individual-level exposure trajectories, and exposure trajectories to both health behavior and longer-run ideological drift.