Theory & Research Mar 27, 2026 18 min read

The Swarm: How Coordinated Digital Manipulation Manufactures Reality at Scale

#blog.tag.d2 #blog.tag.manipulation #blog.tag.propaganda #blog.tag.digital #blog.tag.coordination #blog.tag.bok #blog.tag.encyclopedia #blog.tag.deep-dive

In 2016, researchers at Oxford's Computational Propaganda Project documented state-sponsored social media manipulation in nine countries. By 2023, that number had risen to over eighty. The techniques had evolved from crude bot networks posting identical messages to sophisticated operations involving real humans, AI-generated content, cross-platform narrative laundering, and the strategic exploitation of platform algorithms. Welcome to the age of coordinated digital manipulation — where the line between authentic public discourse and manufactured reality has become nearly invisible. TellDear's Dimension 2 (Manipulation & Propaganda) catalogues nearly a hundred manipulation tactics. While our companion article Manufacturing Reality examines classical propaganda mechanisms and The Invisible Cage explores interpersonal manipulation, this article maps the distinctly modern phenomenon of coordinated inauthentic behavior — how digital swarms form, how they operate, and how to detect them.

I. The Anatomy of Coordinated Inauthentic Behavior

Coordinated Inauthentic Behavior (CIB) is the umbrella term for organized efforts to manipulate public discourse using deceptive means — fake accounts, bot networks, paid trolls, or real people acting under false pretenses. The key word is coordinated: individual trolls and cranks have always existed, but CIB represents something qualitatively different. It is manipulation as infrastructure.

Facebook's definition — later adopted by many researchers — focuses on the behavior rather than the content: accounts that coordinate to deceive others about who they are and what they are doing, regardless of whether the content they share is true or false. This distinction matters. A CIB operation might share entirely accurate news articles — the manipulation lies not in the content but in the artificial amplification, the simulated consensus, the manufactured impression that thousands of ordinary citizens are independently reaching the same conclusion.

The architecture of a typical CIB operation involves several layers:

Command layer: The strategic direction — what narratives to push, what targets to attack, what timing to use. This may be a state intelligence service, a political campaign, a corporate PR firm, or a ideological network.
Operator layer: The people (and increasingly, AI systems) who manage the accounts, create content, and coordinate actions. In Russia's Internet Research Agency, these were full-time employees working in shifts. In many operations, they are freelancers on gig platforms who may not fully understand the larger operation.
Account layer: The fake personas — bot accounts, sock puppets, compromised real accounts, or "rented" accounts from real users. Sophisticated operations maintain accounts for months or years, building authentic-looking histories before activating them.
Amplification layer: The mechanisms that make the operation's content visible — algorithmic manipulation, hashtag hijacking, cross-posting networks, and the exploitation of platform recommendation systems.

Understanding this architecture is crucial because it reveals why CIB is so difficult to counter. Removing individual accounts addresses only the account layer while the command and operator layers remain intact, ready to deploy new accounts. It is, as one researcher put it, "playing whack-a-mole with an adversary that has an infinite supply of moles."

II. Brigading: The Digital Mob

Brigading is the coordinated invasion of an online space by a group with the intent to disrupt, overwhelm, or manipulate. Unlike organic disagreement — where individuals independently discover and respond to content — brigading involves a signal (explicit or implicit) that directs a group to a target. The signal might be a post in a private channel ("Everyone go downvote this"), a coded hashtag, or simply the social dynamics of an outrage-oriented community that functions as a standing army waiting for targets.

The mechanics of brigading exploit a fundamental asymmetry in online discourse: the cost of attacking is far lower than the cost of defending. A brigade of fifty people can each spend thirty seconds leaving a hostile comment on a post that took someone hours to write. The target faces an impossible choice: spend hours responding to each attack (feeding the brigade), delete or hide the content (achieving the brigade's goal), or endure the onslaught (risking psychological damage and the appearance of losing the argument).

Brigading is closely related to dogpiling, but with an important distinction. Dogpiling can occur organically — a tweet goes viral and thousands of strangers independently pile on criticism. Brigading is coordinated: there is a directing intelligence, even if it is just a community's culture of targeting perceived enemies. The distinction matters for analysis because brigading reveals organized intent while dogpiling may emerge from the dynamics of platform virality alone.

Both tactics share a devastating psychological impact. Research on online harassment shows that coordinated attacks produce effects comparable to real-world mobbing: anxiety, depression, self-censorship, and withdrawal from public discourse. This last effect — the chilling of speech — is often the primary goal. Brigading does not need to win arguments. It needs to make certain arguments too costly to make.

III. Sealioning and Gish Gallop: Weaponized Engagement

Not all coordinated manipulation looks like aggression. Some of the most effective tactics disguise themselves as reasonable engagement. Sealioning — named after a 2014 webcomic by David Malki — is the practice of pursuing someone with persistent, superficially polite requests for evidence, explanation, or debate. The sealion frames itself as merely "asking questions" or "wanting to understand," but the true purpose is exhaustion and delegitimization.

Sealioning exploits a norm of good-faith discourse: the expectation that sincere questions deserve sincere answers. The sealion weaponizes this norm by generating an endless stream of questions, each individually reasonable but collectively designed to consume the target's time and energy. Every answer produces three new questions. Every refusal to answer is framed as evidence that the target cannot defend their position. The sealion never concedes a point, never updates their view, never acknowledges a valid response — because the goal was never understanding.

When coordinated, sealioning becomes devastating. Imagine a scientist posting about climate change and receiving dozens of "polite" questions from different accounts: "Can you explain the methodology of that study?" "What about the medieval warm period?" "Could you share the raw data?" "I'm just trying to understand — why do you dismiss solar cycles?" Each question, individually, looks reasonable. Together, they constitute an impossible demand for the scientist to personally tutor dozens of strangers on decades of research — while the questioners invest seconds in each query. This connects directly to the concept examined in our article on Discourse Sabotage, where JAQing (Just Asking Questions) serves a similar function of disguising bad-faith engagement as intellectual curiosity.

The Gish Gallop — named after creationist debater Duane Gish — takes a complementary approach: instead of demanding endless answers, it provides endless claims. The galloper floods a discussion with a rapid series of arguments, each requiring significant effort to refute. The mathematical asymmetry is brutal: it takes ten seconds to make a false claim and ten minutes to properly debunk it. A galloper can produce twenty claims in the time it takes to refute one.

In digital spaces, the Gish Gallop has found its perfect medium. A single comment can contain a dozen hyperlinks to dubious sources, each requiring separate investigation. A coordinated operation can deploy multiple gallopers simultaneously, creating the impression of a well-supported argument through sheer volume. When combined with brigading — where the brigade upvotes the galloped content and downvotes careful rebuttals — the effect is a complete inversion of epistemic quality: the worst arguments appear to be winning.

IV. Manufacturing Consensus and Outrage

Manufactured consensus is perhaps the most fundamental goal of coordinated digital manipulation. The human mind uses the perceived beliefs of others as a heuristic for truth — this is not irrationality but a reasonable adaptation to a world where we cannot verify everything ourselves. If many independent observers report the same thing, it probably is the case. Manufactured consensus exploits this heuristic by simulating independent observation.

The techniques are varied. Bot networks amplify specific messages, making them appear more popular than they are. Astroturfing operations — already covered in our article Manufacturing Reality via the aspect of astroturfing — create the appearance of grassroots support. Sock puppet accounts engage in "conversations" with each other, modeling the consensus they want to create. Paid reviews, fake testimonials, and coordinated rating manipulation all serve the same purpose: making the manufactured appear organic.

The digital amplification of these techniques has produced a qualitative shift. In the pre-digital era, manufacturing consensus required significant resources — printing pamphlets, organizing rallies, buying media coverage. Today, a single operator with fifty accounts can simulate a community. An AI system can generate thousands of unique-seeming comments in minutes. The cost of simulating consensus has collapsed while the psychological power of perceived consensus remains unchanged.

Manufactured outrage is the complementary tactic: instead of simulating agreement, it simulates anger. An operation identifies a potentially divisive issue, amplifies the most extreme positions, and creates the impression that "everyone is furious about this." The goal is not to inform but to activate — to convert passive observers into participants in a conflict that may be largely artificial.

The mechanism relies on what researchers call the "outrage amplification cycle." A provocative claim is seeded by the operation. Bot networks and coordinated accounts amplify it. Platform algorithms, optimized for engagement, further boost the outrage-inducing content. Real users, seeing what appears to be widespread anger, add their own genuine outrage. The manufactured spark ignites real fire. By the time the story peaks, the operation's accounts are a tiny fraction of the total discussion — but they lit the match. This pattern intersects with concepts explored in The Tribal Mind, where ingroup bias and outgroup homogeneity bias make us especially susceptible to narratives of "them versus us."

V. Information Laundering: The Conspiracy to Credibility Pipeline

Information laundering is the process by which false or misleading claims are passed through a chain of increasingly credible sources until they appear legitimate. Like financial money laundering — where illicit funds are moved through legitimate businesses to obscure their origin — information laundering obscures the origin of a narrative by routing it through intermediaries.

A typical information laundering chain might work as follows: A state intelligence service plants a story in an obscure foreign-language blog. A "news aggregator" site picks up the blog post. A fringe English-language outlet reports on the aggregator's story. A partisan commentator on a mainstream platform references the fringe outlet. A mainstream journalist, seeing multiple "sources" reporting the story, considers it worth investigating. At each step, the narrative gains apparent credibility while the original source becomes more distant and harder to identify.

The digital ecosystem has created new and more efficient laundering channels. Social media allows for rapid, low-cost placement of narratives. "Citizen journalism" platforms make it easy to create apparently independent sources. Cross-platform sharing creates the illusion of multiple independent verifications. And the sheer speed of the digital news cycle means that by the time a laundered story is debunked, it has already been cited by dozens of legitimate outlets.

Narrative laundering extends this concept beyond individual claims to entire frameworks of understanding. Instead of laundering a specific false story, narrative laundering works to legitimize a way of thinking — a frame through which events are interpreted. A narrative might be laundered through academic-looking publications, think tanks with legitimate-sounding names, conferences that mix real scholars with operatives, and media appearances that present the laundered narrative as one legitimate perspective among several. The connection to false balance — explored in our article The Symmetry Trap — is direct: once a narrative has been sufficiently laundered, media norms of "presenting both sides" do the rest of the work.

VI. Consensus Cracking: Dividing to Conquer

While manufactured consensus builds false agreement, consensus cracking destroys real agreement. The tactic works by infiltrating communities that share a consensus and introducing doubt, division, and internal conflict. The goal is not to change the community's mind but to fracture its ability to act collectively.

The playbook is well-documented from leaked operations. Operatives identify existing tensions within a target community — and every community has them. They amplify the most divisive voices. They create fake accounts that push extreme positions, making moderates uncomfortable. They introduce purity tests that exclude allies. They weaponize procedural disagreements to consume the community's energy in internal disputes rather than external action.

A classic consensus-cracking operation against an environmental organization might involve: creating accounts that push extreme positions ("anyone who drives a car is the enemy"), accounts that amplify internal disagreements ("the leadership doesn't represent the grassroots"), accounts that introduce whataboutism directed inward ("why aren't we talking about [unrelated issue] instead?"), and accounts that sealion community leaders with endless procedural challenges. None of these interventions are obviously hostile. Each could be mistaken for genuine community discourse. Together, they are lethal to collective action.

The concept has deep roots in intelligence operations — the FBI's COINTELPRO program used strikingly similar tactics against civil rights organizations in the 1960s. Digital platforms have simply made these operations cheaper, faster, and harder to detect. A single operator can simultaneously pose as a moderate voice of reason in one thread and a radical provocateur in another, playing both sides against the middle.

VII. The Platform Amplification Problem

Coordinated digital manipulation does not operate in a vacuum. It operates within platform architectures that were designed to maximize engagement — and nothing engages like conflict, outrage, and the feeling of being part of a crowd. This creates a structural amplification effect: platforms are optimized to spread exactly the kind of content that CIB operations produce.

Consider the algorithmic incentives. A brigading attack generates massive engagement metrics — comments, shares, reactions. Platform algorithms interpret this as "interesting content" and boost it to wider audiences. Manufactured outrage triggers emotional responses that increase time-on-platform. Gish gallops generate lengthy threads that count as "deep engagement." The platform's own systems become unwitting force multipliers for the manipulation.

This is not a bug that can be patched. It is a fundamental tension between platform business models (which depend on engagement) and information integrity (which requires the suppression of manipulative engagement). As long as platforms are optimized for engagement metrics, they will be structurally vulnerable to operations that generate artificial engagement. The manipulators are, in a sense, the platforms' best users — they generate exactly the behavior the algorithms are designed to promote.

The recommendation systems add another layer. A user who engages with content targeted by a brigading attack will be recommended similar content, creating a feedback loop. A user who shares manufactured outrage will be shown more outrage-inducing content. The platform's personalization systems, designed to give users "more of what they want," become delivery mechanisms for manipulation — connecting the willing audience with the manufactured message, as if by invisible hand.

VIII. Detection: Reading the Patterns

If coordinated digital manipulation is designed to be invisible, how can ordinary citizens detect it? While no method is foolproof, several patterns provide reliable signals.

Temporal patterns: Organic discourse develops gradually. Coordinated operations often show sharp, synchronized spikes in activity. If a hashtag goes from zero to trending in minutes, with most posts coming from a narrow time window, coordination is likely. Real grassroots movements build over hours and days, not minutes.

Account patterns: Bot and sock puppet accounts often share telltale characteristics — recent creation dates, generic profile photos (or AI-generated faces with telltale artifacts), low follower-to-following ratios, and posting patterns that suggest automated scheduling (posts at exactly regular intervals, activity at times inconsistent with any single time zone).

Content patterns: Coordinated operations often share identical or near-identical talking points. While real communities develop shared language organically, coordinated accounts often use suspiciously similar phrasing — because they are working from the same script. The phenomenon of dozens of "independent" accounts posting nearly identical comments is a strong signal of coordination.

Network patterns: Mapping who shares whose content reveals coordination structures. Organic networks show messy, overlapping clusters. Coordinated networks often show unusually tight reciprocal sharing patterns — accounts that always amplify each other but rarely interact with the broader community.

Emotional patterns: Manufactured content tends to be emotionally extreme and binary. Real discourse contains nuance, uncertainty, and qualification. If every voice in a "debate" expresses absolute certainty and maximum outrage, the debate may not be real. This connects to the bias blind spot discussed in The Mirrors of Self-Deception — we are often better at detecting manipulation aimed at others than at recognizing when we ourselves are being manipulated.

IX. Case Studies: The Swarm in Action

The Internet Research Agency (2014–present)

Russia's Internet Research Agency (IRA) represents perhaps the best-documented example of state-sponsored CIB. Operating from an office building in St. Petersburg, the IRA employed hundreds of people to operate fake social media accounts targeting American, European, and Russian audiences. The operation's sophistication was remarkable: operatives created detailed American personas, organized real-world events (including both pro-Trump and anti-Trump rallies in the same city on the same day), and built genuine followings in niche communities before activating them for political purposes.

The IRA's strategy was not to promote a single narrative but to amplify division on every available axis — race, religion, politics, immigration. Success was measured not by changing minds but by increasing social tension and eroding trust in institutions. By the time the operation was publicly exposed in 2018, it had reached an estimated 126 million Americans through Facebook alone.

Commercial Manipulation

State actors get the headlines, but commercial CIB may be more pervasive. The "review manipulation" industry — fake product reviews, fake restaurant ratings, fake app store reviews — represents coordinated inauthentic behavior at massive scale. Companies like Devumi sold fake Twitter followers by the million. Click farms in developing countries provide engagement-for-hire services. The "influence-for-sale" economy has created a marketplace where anyone can buy the appearance of consensus.

This commercial dimension deserves attention because it normalizes the very tactics that political operations exploit. When buying fake reviews is a routine business expense, and when everyone "knows" that online ratings are gamed, the cultural antibodies against manufactured consensus are weakened. We learn to distrust online signals without developing better tools for evaluating claims — leaving us more vulnerable, not less, to sophisticated manipulation.

Health Misinformation Swarms

The COVID-19 pandemic provided a real-time laboratory for studying coordinated manipulation. Research identified organized networks spreading anti-vaccine content, miracle cure claims, and conspiracy theories. These networks combined automated amplification (bots), coordinated human operators, and the exploitation of genuine public uncertainty. The "Disinformation Dozen" — twelve individuals responsible for an estimated 65% of anti-vaccine misinformation on social media — demonstrated how a small coordinated core could leverage platform dynamics to reach billions.

X. Defenses and Their Limits

Technical countermeasures — bot detection, network analysis, content moderation — are necessary but insufficient. They address symptoms while the structural incentives for manipulation remain intact. Several deeper approaches deserve consideration.

Structural literacy: Rather than teaching people to identify specific manipulations (which evolve faster than training), we should build understanding of the structural dynamics that make manipulation effective. Why does manufactured consensus work? Because we use social proof as a heuristic. Understanding the heuristic makes us better at questioning it when appropriate.

Friction by design: Platforms could introduce deliberate friction — slowing the spread of content that shows coordination signals, requiring verification for accounts that reach certain amplification thresholds, and making it harder to create accounts at scale. Some friction is already being introduced (Twitter/X's verification changes, Meta's labeling of state-affiliated media), but these measures remain modest relative to the problem.

Inoculation: Research on "prebunking" — exposing people to weakened forms of manipulation so they develop resistance — shows promising results. When people understand the techniques of manufactured consensus before encountering them, they are significantly less susceptible. This is precisely the mission of critical thinking education platforms like TellDear: not to tell people what to think, but to make the mechanisms of manipulation visible.

Collective defense: Individual vigilance is important but insufficient against coordinated operations. Effective defense requires collective capacity — communities that can identify, discuss, and respond to manipulation collectively. Ironically, this requires exactly the kind of social trust that coordinated manipulation is designed to destroy — which is precisely why consensus cracking is among the most strategically important tactics in the swarm's repertoire.

XI. The Epistemological Crisis

The deepest challenge posed by coordinated digital manipulation is not any specific campaign but the cumulative effect on our collective ability to know things. When manufactured consensus is cheap and easy, the signal value of apparent agreement is degraded. When information laundering is routine, the signal value of institutional endorsement is degraded. When sealioning and brigading make public discourse costly and unpleasant, the people most qualified to contribute withdraw — leaving the field to the loudest and the most coordinated.

This is the reverse cargo cult effect at civilizational scale: if nothing can be trusted, then the manipulators and the truth-tellers are on equal footing. "Everyone lies" is not a defense against propaganda — it is propaganda's ultimate victory condition. When a society reaches the point where citizens believe that all information is manipulated, they do not become better critical thinkers. They become nihilists — and nihilists are the most manipulable audience of all, because they have abandoned the very concept of evidence-based reasoning.

Recognizing the mechanics of coordinated manipulation — as catalogued in TellDear's Dimension 2 — is therefore not just an academic exercise. It is a survival skill for democratic societies. The swarm is not going away. But understanding its structure, tactics, and vulnerabilities is the first step toward building a discourse ecosystem that is resilient against it. The alternative — a public sphere where manufactured reality is indistinguishable from authentic expression — is a future worth fighting to prevent.