# Troll Farms A coordinated industrial-scale operation — physical office, paid staff, shift work, scripted personae, performance metrics, and multi-account workflows — producing high volumes of social-media content designed to amplify particular narratives, harass particular targets, simulate grassroots support for particular positions, or polarize public discourse around chosen wedges. The defining feature is *coordination at scale*: not individuals trolling on their own time but an assembly-line operation with managers, KPIs, and contractor pipelines. ## The Saint Petersburg template: the IRA The contemporary use of *troll farm* in English is essentially backshifted from Russian-language journalism describing the [[Internet Research Agency]] (IRA, *Агентство интернет-исследований*) — a [[Yevgeny Prigozhin]]-funded operation that occupied an office at 55 Savushkina Street in Saint Petersburg from 2013 until its public dissolution in 2023 (post-Prigozhin mutiny). The IRA template, documented by Russian journalist Lyudmila Savchuk (who infiltrated the operation as a freelance writer in 2015 and successfully sued IRA in Russian court for unpaid wages, producing the first detailed insider account), included: - **Twelve-hour shifts** with staff working in teams on assigned platforms (LiveJournal, VKontakte, Twitter, Facebook, [[YouTube]]) - **Daily quotas** — measured comments, posts, shares per shift - **Scripted personae** — staff worked from "legends" describing the imagined American or European persona they were posting as; English-language operators used culturally-researched personae targeted at specific US audience segments - **Topical guidance** distributed daily — which narratives to amplify, which to attack, which targets to harass - **Multi-account workflows** — single operators managing 5–20 personas across platforms - **Cross-platform amplification chains** — content seeded on one platform, screenshotted and shared on another, then amplified by automated bot accounts The February 16, 2018 Mueller indictment of 13 IRA staff plus three companies (Concord Management, [[Concord Catering]], and the IRA itself) provided the first US-criminal-prosecution evidentiary record of the operation. The [[Senate Intelligence Committee]]'s bipartisan 2019–2020 reports (especially Volume 2: Russia's Use of Social Media) provided the most-comprehensive documentation of IRA activity targeting US audiences from 2014–2017. ## Beyond Russia: the proliferation By the late 2010s the troll-farm template had been adopted by other state and state-adjacent actors: - **China** — the *Wumao Party* (50-cent-army; commenters allegedly paid 0.5 RMB per pro-government post; later studies suggest the operation is more often unpaid CCP staff than freelance contractors) - **Iran** — multiple Endless Mayfly / Liberty Front Press operations targeting English-language audiences with anti-Saudi, anti-Israeli, anti-US content; Facebook and Twitter takedowns 2018–2020 - **India** — the BJP's IT Cell documented by *The Print*, *The Caravan*, *The Wire* as a national-scale Hindi-and-English content operation - **[[Saudi Arabia]]** — the post-Khashoggi-killing operations on Twitter; Twitter Saudi-aligned-account purges - **Philippines** — extensive documentation by Rappler / Maria Ressa of Duterte-era troll networks - **Brazil** — the *Gabinete do Ódio* (Hate Office) operating from inside the Bolsonaro Planalto Palace, documented by the Congressional COVID-19 CPI inquiry - **Mexico** — *Peñabots* documented as early as the 2012 cycle - **[[Belarus]], Hungary, Turkey, Venezuela** — variants of the template with state-aligned content operations Commercial / for-hire versions also proliferated. The [[Cambridge Analytica]] / SCL Group operations included content-production components; Israeli firms (Black Cube, Archimedes Group / Aviv Group, Psy-Group) and various Western for-hire reputation-management agencies offer similar service stacks under more genteel labels. ## What troll farms actually do (the work product) The output of a troll farm typically falls into several content categories: 1. **Amplification** — boosting existing native content from real political figures with massed comments, shares, retweets 2. **Wedge content** — original content engineered for specific demographic-targeted polarization (BLM-themed content aimed at Black audiences and counter-content aimed at white-conservative audiences, run from the same operation, designed to deepen the wedge) 3. **Issue laundering** — placing fringe narratives in low-trust outlets, then citing those outlets in higher-trust venues until the narrative is "in the conversation" 4. **Harassment campaigns** — coordinated targeting of named individuals (journalists, dissidents, opposition figures) with abuse, doxxing, and pile-on reporting 5. **Astroturf event organizing** — creating fake event pages, recruiting real attendees, occasionally generating real protests from synthetic seed organizing (the IRA documented as having organized actual US protests in 2016 by both pro- and anti-Trump audiences) 6. **Counter-narrative seeding** during crisis events — flooding hashtags with off-topic or distracting content to drown out organic conversation ## Why the scale matters Two analytical claims about troll-farm operations are worth keeping distinct: 1. **The direct persuasion effect is modest** — the academic literature on social-media persuasion is consistent that any particular post or even any particular campaign rarely changes minds at scale. The IRA's documented 2016 reach was measurable but small relative to organic political content 2. **The structural effect is larger** — sustained troll-farm operation degrades the *epistemic infrastructure* of platforms (everyone learns to be suspicious of strangers, organic conversation gets harder), provides a *cover story* for organic toxic content ("that wasn't real Americans, that was Russians"), and *trains political consultants* in techniques that then bleed into legitimate domestic political work The strategic question for the 2020s — especially in the LLM era — is no longer "is there a troll-farm operation targeting us?" (the answer is reliably yes from multiple state actors) but "what are the cumulative effects on public discourse of normalized industrial content operations on every major platform?" ## See Also - [[Internet Research Agency]] — the canonical operation - [[Yevgeny Prigozhin]] — the IRA funder - [[Sockpuppet Accounts]] — the account-level building block of troll-farm work - [[Astroturfing]] — the broader category of manufactured grassroots - [[Active Measures]] — the Russian-intelligence parent doctrine - [[Doxxing]] — the most-frequent troll-farm content category for harassment campaigns - [[Microtargeting]] — the data-targeting layer often paired with troll-farm content production - [[Disinformation Hub]] — Subject hub - [[Disinformation Lexicon]] — alphabetical MOC