From facts to narratives: rethinking what disinformation is
A through-line in this collection is dissatisfaction with the dominant operationalization of disinformation as discrete false claims attached to low-credibility domains. Marwick2026-qd reconceptualizes disinformation as cross-platform, identity-affirming cathartic narrative; Sadler2025-vu argues from hermeneutic realism that ethically problematic stories can be referentially accurate; Prochaska2025-ef and Starbird2025-jj show how “deep stories” and evidence–frame interactions on Twitter make individual posts misleading only in the context of collective sensemaking. Suau_Martinez2026-lv and Waight2025-al take this narrative turn empirically — the former through survey evidence that cumulative narrative exposure increases belief, the latter through an LLM-based method for measuring cross-language narrative similarity. Goel2025-iq sharpens the point: mainstream articles co-shared with fake news are demonstrably repurposed to support misleading narratives, meaning that domain-list approaches systematically undercount the phenomenon. Giglietto2019-e9be81c1 anticipates much of this by shifting analytic focus from creators’ intent to propagators’ judgments within a hybrid news system. Several conceptual interventions extend the critique further: Frischlich2025-vn argues that virus, warfare, and pollution analogies each obscure parts of a multilevel adaptive system; Dierickx2026-tw proposes “emergent facts” as a new epistemic category for generative AI outputs; Yoo2026-ev shows how “fake news” itself has become a weaponized label rather than a descriptor.
Measurement, methods, and the WEIRD-centric gaze
A second strand is methodological self-scrutiny. Luhring2025-od audits NewsGuard, showing that binary trustworthy/untrustworthy thresholds can distort findings dramatically and that source coverage skews US-ward. Rossi2023-847d5a9f extends Guess et al.’s URL-shares approach to France, Germany, and Italy, while Hourigan2026-oc uses digital diaries to let participants flag misinformation without imposed definitions — finding business/economics content and clickbait mainstream headlines flagged far more often than political fringe sources. Humprecht2025-ml and Nenno2025-xa explicitly press for de-WEIRD-ing the field; Ventura2025-sw does so by running a WhatsApp deactivation experiment in Brazil, Rossini2026-jn by linking electoral misinformation to political intolerance there, and Gaw2025-ru by theorizing Philippine influence operations as brokerage. Mosleh2024-op insists on cross-platform comparison beyond the Twitter/Facebook duopoly. Scalco2026-bd operationalizes “information voids” quantitatively via demand–supply imbalances; Waight2025-al and DeVerna2025-dl demonstrate the promise (and ideological skew) of LLMs in the measurement pipeline. Mahl2026-hc synthesizes these concerns into a Delphi-based anticipatory agenda.
How big is the problem, and where does it actually live?
The collection contains a productive tension between alarmist and deflationary accounts. Budak2024-ef argues public discourse overstates average exposure, algorithmic causation, and societal harm. Consistent with this, Lyons2026-ca finds low-credibility health exposure rare and concentrated among older adults; Vincent_undated-re flags TikTok as a relative outlier in European measurements. Yet other studies recover sizable downstream effects when looking at the right targets: Bollenbacher2026-vz estimates antivaccine tweets causally drove ~14,000 refusals and excess deaths; Kim2026-wg shows that targeted digital voter suppression depressed turnout substantially among the very subpopulations average treatment effects obscure. Gonzalez-Bailon2024-rq and Renault2025-uh document concentrated viral diffusion and partisan asymmetries on Facebook and X respectively. Adam2026-tz integrates these positions by showing both contagion (via alternative media) and mitigation (via mainstream debunking), conditional on predispositions.
Producers, brokers, and infrastructures of manipulation
A cluster of papers opens up the supply side. Poliakoff2026-fa reconstructs the IRA as a Prigozhin-owned PR-style firm embedded in Russia’s media labor market, complicating the “state-funded troll factory” image. Thiele2026-or… actually Thiele2025-ol proposes a rational-choice typology of coordinated social media manipulation, Gaw2025-ru theorizes influence operations as brokerage, and Graham2025-gp shows how propaganda exploits rather than evades the infrastructures of truth. Detection and coordination work — Rodriguez_Farres2025-sg on Bluesky, Zhao2025-ny on hashtag hijacking, Kuznetsova2025-nu on pro-regime Telegram in Russia/Belarus, Pante2025-pq on inter-state coordination claims, Kim2026-br on Korean trolls, and Arceneaux2026-xk on bots as agenda-builders — collectively reframe disinformation as organized labor and computational infrastructure rather than emergent error. Copland2025-em situates Sky News Australia within transnational network propaganda, while Donovan2025-ws argues misinformation-at-scale is a structural feature of engagement-driven business models. FitzGerald2025-nv highlights the temporal adaptivity of such campaigns.
Movements, mainstreaming, and the radical right
Several papers trace how protest, party, and platform combine to mainstream extreme content. Rothut2026-or develops “protest-facilitated mainstreaming” via Querdenken on Telegram; Bailard2024-pj links Proud Boys’ framing dynamics to offline violence; Bennett2025-xs theorizes “digital surrogate organizations” pulling parties toward illiberalism; and Tornberg2025-ir shows cross-nationally that it is specifically radical-right populism — not populism or right-wing ideology alone — that predicts elite misinformation sharing. Askanius2026-de maps the hybridized Swedish far-right YouTube influence network; Rieder2026-pp documents how the “Tate-space” survives deplatforming through ambient ideology and remix; Marwick2025-vx reframes “redpilling” as processual socialization rather than sudden conversion; and Marwick2025-ov complicates the demographic stereotype of the conspiracy believer through TikTok’s “populist knowledge production.” Gattermann2025-yx connects this back to mass attitudes by linking far-right electoral performance to disinformation concern.
Conspiracy, belief formation, and audience perceptions
Several studies investigate how beliefs form and what citizens themselves perceive. Adam2026-tz, Rohrbach2026-rc, and Marwick2025-ov each emphasize that mainstream journalism and platform vernaculars — not just fringe content — shape conspiratorial uptake. Van_Erkel2026-mk introduces the “hostile misinformation effect” — citizens believe their own side is the disproportionate target — paralleling the classic hostile media effect. Choi2026-bz traces modality-congruent carryover effects of (deep)fake encounters on metacognitive confidence, while Hameleers2026-mc finds that AI-generated visuals are not uniformly more persuasive than decontextualized real video. Lieu2025-nl shows that content categories (especially solutions denial) matter more than fallacy types in climate misinformation, and Gardam2025-er argues for visual-first methodologies attuned to platform multimodality. Bosch2024-hj adds sound as an underappreciated propaganda affordance on TikTok.
Interventions: fact-checking, prebunking, inoculation, and their limits
The collection’s intervention literature is internally conflicted. van-der-Linden2026-jt and Szabo2026-rd demonstrate that prebunking and conversational inoculation can durably increase resistance at scale; Costello2024-bg reports striking durable reductions in conspiracy belief through tailored AI dialogue. Dubey2026-bl finds even high-conspiracy users accept balanced news chatbots. Yet Cazzamatta2026-lo shows fact-checkers themselves favor labels and counter-speech over removal — and are now positioned as “censors” by Meta’s policy shift — while Farkas2026-lr documents their rhetorical struggle to legitimize platform partnerships. DeVerna2025-dl finds that even reasoning LLMs need curated RAG to fact-check reliably, and Spampatti2026-kx argues from climate psychology that information provision alone is insufficient. Xue2025-bp complicates the objectivity self-image of fact-checking by foregrounding its emotional register. Simeone2025-vo makes the case that targeted deplatforming can be highly effective at disrupting authority structures, even as Rieder2026-pp shows ambient ideologies survive moderation. Tai2026-qk reminds us that institutional and ideological positioning of elites themselves conditions misinformation supply.
Generative AI, platform governance, and the receding horizon of truth
A final, increasingly urgent thread concerns generative AI and platform restructuring. Emilio2026-ik articulates the “Generative AI Paradox” — synthetic realities may rationally lead users to discount all digital evidence — and Waight2026-ts shows how state-controlled media is already laundered through LLMs’ training data and outputs. Triedman2025-uy empirically audits Grokipedia, finding heavy derivation from Wikipedia with systematic injection of low-quality and blacklisted sources on controversial topics. Dierickx2026-tw and Choi2026-bz probe the epistemic and psychological consequences. Alongside these AI-centric pieces sit broader diagnoses of platform decline and capture: Bruns2026-yv charts Twitter’s “death” and successor platforms’ inadequacies; Cazzamatta2026-lo and Renault2025-uh dissect Meta’s and X’s retreats from professional fact-checking; Di-Domenico2026-zq traces how influencer credibility itself becomes an infrastructure for toxic misinformation. Read together, these papers describe a hybrid media ecology in which the locus of the problem has migrated from discrete falsehoods to the very infrastructures of attention, authority, and verification — a migration that the field’s narrative, infrastructural, and complexity-oriented reframings are now scrambling to follow.