From Detection Tactic to Conceptual Framework
The papers gathered here trace the trajectory of coordinated inauthentic behaviour (CIB) from a Facebook policy term into a sustained research programme. The foundational move is made in Giglietto2020-9d8acdd7, which reframes disinformation research away from content veracity and bad-actor taxonomies toward the detection of coordinated link sharing behaviour (CLSB) during the 2018 and 2019 Italian elections. This ecological, behaviour-centred reframing — operationalised through the CooRnet package — is then extended into the pandemic in Giglietto2022-0e951ac5, where CLSB becomes a real-time tool for fact-checkers, and updated in Giglietto2023-fa71a001 with a workflow for continuously refreshing coordinated-account lists, addressing the perishability problem that plagues static blocklists. The most recent instalment, Giglietto2026-9b6a992d, turns the same lens on coordinated gambling promotion, documenting how generative AI intensifies rather than replaces established persuasion drivers within already-coordinated networks.
Methodological Diversification
A second cluster of work treats the methodological frontier: how to detect coordination more robustly. Yang2025-iv addresses the central vulnerability of timing-based detection by exploiting statistical regularities in sharing speed and frequency. Mannocci2025-ig and Iannucci2025-eg both push toward multimodal and multiplex temporal representations, arguing — with empirical support on UK election and benchmark datasets — that no single co-action modality dominates and that flattening layers discards structurally important nodes. Gerard2025-br pushes further still, with CANE/t-CANE replacing interaction graphs entirely with affiliation networks over latent narrative clusters, a move motivated partly by the collapse of API access. Foundation-model approaches appear in Minici2024-tf, while Bastos2025-ol opens a visual-identity channel for attribution. Platform-specific adaptations extend the toolkit to TikTok Luceri2025-tr, Bluesky Rodriguez_Farres2025-sg, WhatsApp Kansaon2025-id, Telegram Kuznetsova2025-nu, and Reddit Graham2026-fb — the last reframing ‘botting’ as participatory culture and challenging the human/bot binary that underwrites most detection work.
Conceptual Critique and Reconstruction
Running alongside the methodological expansion is a sharpening critique of the CIB concept itself. Thiele2025-ol argues that ‘inauthenticity’ conflates a normatively secondary feature (deception) with the genuinely problematic one (manipulative intent), and proposes ‘coordinated social media manipulation’ grounded in rational-choice attribution. Hurcombe2025-cs performs the complementary discourse-analytic move, showing how Meta’s Newsroom has used ‘coordinated inauthentic behaviour’ as a floating signifier that locates problems outside the platform’s own affordances and business model — a framing that has since diffused into academic and regulatory vocabularies. Graham2026-fb adds an ethnographic challenge: authenticity is a community-negotiated performance, not a technical property. Together these papers ask whether the field has inherited a policy-friendly category that now constrains its own analytical horizons.
Re-evaluating Impact and Reach
A striking convergence in recent work questions whether coordinated networks matter as much as the literature implies. The large-scale FIES analyses Appel2026-qr and noauthor_undated-bm show that Meta-identified deceptive networks reached substantial audiences but that exposure was extraordinarily concentrated, mostly indirect via reshares from non-network users, and — crucially — that associations with downstream political outcomes vanish under proper covariate adjustment. Di-Marco2025-aa reaches a parallel conclusion from a graph-theoretic angle: the placement of coordinated accounts in real cascades resembles random more than strategic positioning, producing roughly an order of magnitude less influence than optimal benchmarks. Pante2025-pq applies the same scepticism to claims of inter-state coordination, arguing prior findings did not survive modern multi-signal detection with controls. Goel2025-iq reframes the problem entirely: the more consequential phenomenon may be the co-sharing of mainstream content to support misleading narratives, a dynamic invisible to domain-list and CIB-focused measurement.
Contexts of Use: Politics, Health, Authoritarianism
A third strand applies CIB detection to substantive domains. Song2025-yh uses CLSB to compare pro- and anti-vaccine coordination in the UK and US, finding — importantly — that coordination is ideologically agnostic and can amplify both credible public-health messaging and misinformation. Efstratiou2026-ij traces information pathways from coordinated contrarian retweet networks through credentialed superspreaders to low-trust outlets. Electoral applications include Oprea2025-lf on Romanian European Parliament campaigns, Kansaon2025-id on Brazilian WhatsApp mobilisation toward the 8 January 2023 attacks, and FitzGerald2025-nv on the appropriation of emerging events to sustain campaigns over time. Authoritarian and geopolitical contexts foreground a different politics: Kuznetsova2025-nu on Russian and Belarusian pro-regime Telegram amplification, Jovanovic-Harrington2026-ze on the mimetic, decentralised CIB of Serbia’s SNS, Kulichkina2026-zk on the dual-use coordination around China’s COVID-19 protests, Graham2025-gp on IStandWithPutin’s parasitic exploitation of platform infrastructures, Zhao2025-ny on hashtag hijacking, and Kim2026-br on two decades of suspected troll activity in Korean news comments, where condemning rhetoric outperforms direct praise.
The Generative-AI Horizon
The most recent papers register a shift in the threat model. Orlando2025-ul uses generative agent-based modelling to show that mere teammate awareness among LLM agents produces coordination dynamics nearly equivalent to explicit deliberation — emergent CIB without a principal. Schroeder2026-im generalises this into a policy argument about malicious AI swarms, while Giglietto2026-9b6a992d documents the empirical correlate: a 13,000% post-volume increase in coordinated gambling networks after ChatGPT’s release, with AI aesthetics recombining established persuasion drivers. These papers suggest the field’s next phase will need to address coordination that is emergent rather than principal-directed, and persuasion infrastructures that scale without the human labour bottleneck that shaped a decade of CIB research.
An Arc of Inquiry
Read together, the collection traces a recognisable arc. An initial methodological breakthrough — behavioural detection of coordination via CLSB and CooRnet — generated a productive research programme that broadened across modalities, platforms, and domains. As the programme matured, it turned reflexive: questioning its inherited vocabulary Thiele2025-ol Hurcombe2025-cs, its binary ontologies Graham2026-fb, its overstated effects Appel2026-qr Di-Marco2025-aa Pante2025-pq, and its narrow attention to overtly inauthentic actors when mainstream-content repurposing Goel2025-iq and emergent agentic coordination Orlando2025-ul may matter more. The field’s open question is whether ‘coordinated inauthentic behaviour’, as both concept and detection paradigm, can be reconstructed to accommodate this expanded problem space — or whether, as several of these papers hint, it must give way to a successor framework organised around manipulation, amplification infrastructure, and emergent coordination rather than inauthenticity per se.