In 1922, the journalist Walter Lippmann published Public Opinion, the foundational text of modern media theory, and gave one of its chapters a title that would outlive the book: "The Manufacture of Consent." Lippmann did not mean the phrase as an accusation. He meant it as a description of a necessary art. The modern public, he argued, was too large, too distracted, and too poorly informed to govern itself directly; it lived not in the world but in a "pseudo-environment" of images planted in its head by others. Someone had to manage those images. "The manufacture of consent," Lippmann wrote, "is capable of great refinements... and the process by which public opinion arises is certainly no less intricate than it has appeared in these pages, and the opportunities for manipulation open to anyone who understands the process are plain enough." A "specialized class" would do the understanding; the rest — what Lippmann would later call the "bewildered herd" — would be managed for their own good. The young Edward Bernays & The Engineering of Consent, Freud's nephew, read Lippmann and turned the diagnosis into a profession.
That the manufacture could be done at scale had just been demonstrated. During the First World War, the Wilson administration's Committee on Public Information — the Creel Committee, in whose orbit both Lippmann and the young Edward Bernays & The Engineering of Consent worked — had taken an American public overwhelmingly opposed to entering a European war and, within months, converted it into a furnace of anti-German enthusiasm. Bernays drew the lesson without flinching in his 1928 manual Propaganda: "The conscious and intelligent manipulation of the organized habits and opinions of the masses is an important element in democratic society. Those who manipulate this unseen mechanism of society constitute an invisible government which is the true ruling power of our country." What Bernays sold as a commercial service and Lippmann defended as a civic necessity, Herman and Chomsky would later anatomize as an impersonal structure. All three were describing the same discovery: in a mass democracy, consent is not found. It is produced.
Not everyone in the 1920s accepted Lippmann's prescription. The philosopher John Dewey, in The Public and Its Problems (1927), answered him directly, refusing the premise that the public was a permanently bewildered herd to be managed by experts and insisting that a democratic public could be educated into genuine competence if its institutions were arranged to inform rather than to steer it. The "Lippmann–Dewey debate" — whether democracy means an informed public governing itself or a managed public consenting to its governors — is the argument Herman and Chomsky walked back into sixty years later. Their wager was Dewey's, made empirical: if the public is incompetent, perhaps it is because the institutions that inform it are built to keep it so.
Sixty-six years after Lippmann, the economist Edward S. Herman and the linguist Noam Chomsky borrowed his phrase and inverted his approval. Manufacturing Consent: The Political Economy of the Mass Media (Pantheon, 1988) took the manufacture of consent not as the benign machinery of democracy but as its quiet subversion — the means by which a nominally free press in a market society reliably produces a worldview congenial to the elites who own it, fund it, and supply it with information. The book's central claim was deceptively modest and, to its critics, infuriating: that you do not need censorship, state control, or a conspiracy to explain why the American press marches in step on the questions that matter to power. You need only follow the money and the institutional incentives. The press is free, Herman and Chomsky argued, and that freedom is precisely what makes the result so persuasive — and so invisible.
The proof they offered most often was a pair of corpses. In the late 1970s, two genocides were unfolding simultaneously. In Cambodia, the Khmer Rouge were murdering somewhere near 1.7 million people, and the American press covered the atrocity with saturation outrage. In East Timor, the Indonesian military — invading in December 1975 with weapons that were ninety percent American-supplied, and a green light delivered to Suharto by Gerald Ford and Henry Kissinger the day before the troops landed — was killing a comparable fraction of a far smaller population, perhaps 200,000 people. The New York Times, which had covered East Timor before the invasion, fell nearly silent the moment the killing began. Cambodia's killers were official enemies; the crime could be deployed. East Timor's killers were a strategic ally; the crime had to be unseen. Same years, same scale of horror, opposite coverage. That asymmetry, Herman and Chomsky argued, was not an accident or a failure. It was the system working exactly as designed.
The two authors had personal reason to believe the machinery operates on raw instruction as readily as on incentive. In 1973 they had published Counter-Revolutionary Violence: Bloodbaths in Fact and Propaganda through a small house, Warner Modular Publications, which printed roughly 20,000 copies. An executive at the parent conglomerate, Warner Communications, saw an advertisement for the book, judged its indictment of American conduct in Indochina intolerable, and — rather than distribute it — moved to shut the subsidiary down and destroy the entire run. No government censor was involved. A corporate owner simply declined to circulate a critique of power. The episode became, for Herman and Chomsky, the founding illustration of the thesis they would not formalize for another fifteen years: that in a private media economy the first filter can be exercised by an owner's veto as quietly as a state's, and with no court to appeal to.
The mechanism is a model, not a metaphor. In the book's first chapter, Herman and Chomsky lay out five "filters" through which raw news must pass before it reaches the public, each one screening out what is inconvenient to concentrated power and letting through what is congenial. The filters are structural; they operate whether or not any individual journalist is honest, and most are.
The first filter is concentrated ownership. The dominant media are large, profit-maximizing corporations owned by still-larger conglomerates and wealthy individuals, integrated into the broader corporate economy through interlocking directorates, shared bankers, and common dependence on government licensing and policy. A handful of firms — by the 1980s around fifty, today closer to six — control the overwhelming majority of what Americans read and watch. An institution structured to maximize shareholder return does not need to be told to avoid news that threatens the corporate order; it simply tends not to produce it. This is why Corporate Personhood & The Corporation sits at the base of the model: the medium is itself a corporation, and it reports on the world from inside that skin. The trend the model named has only sharpened. When Ben Bagdikian published The Media Monopoly in 1983, some fifty firms controlled most of what Americans read and watched; by the 2004 edition, retitled The New Media Monopoly, the figure had fallen to five. Concentration on that scale means that editorial judgment about the affairs of the powerful is exercised inside a handful of boardrooms that are themselves among the powerful — General Electric, a major defense contractor, owning NBC; a private-equity fund or a billionaire proprietor setting the budget of the very newsroom that is meant to scrutinize his class.
The second filter is advertising, which Herman and Chomsky call the de facto licensing authority of the press. Newspapers and broadcasters do not primarily sell content to readers; they sell readers — specifically, affluent readers with disposable income — to advertisers, who supply the bulk of revenue. A publication that alienates advertisers dies, regardless of how many readers it has; the radical and working-class press of the early twentieth century collapsed not because it lacked an audience but because it could not attract the advertising subsidy its competitors enjoyed. The result is a built-in tilt away from content that is hostile to business interests or merely "unentertaining" to a buying audience, and toward a tone the advertiser finds safe. The clearest demonstration is a death. In 1964 Britain's Daily Herald, a Labour-aligned paper, folded with a readership of around 4.7 million — nearly double the combined circulation of The Times, the Financial Times, and The Guardian. It died not for want of readers but because its readers were the wrong ones: working-class, low-income, and therefore close to worthless to the advertisers whose subsidy kept its rivals alive. As the historians James Curran and Jean Seaton documented, the advertising filter does not merely color content; it sets the economic threshold below which a viewpoint, however popular, simply cannot pay its way. A market in readers would have kept the Herald alive. A market in audiences-sold-to-advertisers killed it.
The third filter is sourcing. News is produced on a deadline and a budget, and the cheapest, most reliable, most legally defensible source of a constant stream of "newsworthy" material is officialdom — government agencies, the Pentagon, corporate press offices, and the credentialed "experts" those institutions fund and certify. The press stations its reporters where the documents and the briefings are. This creates a symbiotic dependency: the journalist needs access, the official controls access, and access is granted to those who do not betray the relationship. The result is a press that reflexively adopts the framing of its primary sources and treats the powerful as the natural narrators of events. It is the same capture that Operation Mockingbird achieved by directly inserting CIA assets into newsrooms — except here no asset is needed, because the structure of beat reporting produces the deference on its own. The British media scholars Stuart Hall and his colleagues, writing in Policing the Crisis (1978), gave the dynamic a name Herman and Chomsky's filter formalizes: the "primary definers." The accredited officials who speak first and with authority get to set the frame of a story, and the journalist who must respond on deadline is structurally relegated to the role of secondary definer, arguing inside boundaries someone else has already drawn. The reporter may be skeptical, even hostile, and still end up transmitting the official's terms simply because those were the terms the story arrived in.
The fourth filter is "flak" — the negative response a story or outlet provokes when it strays. Flak takes the form of letters, lawsuits, advertiser boycotts, congressional pressure, and organized campaigns by well-funded institutes and pressure groups designed specifically to discipline the media and raise the cost of deviation. Because flak from the powerful is expensive and flak from the powerless is not, the press learns to anticipate and avoid the former. A story that will draw a coordinated assault — the destruction of Operation Mockingbird's Gary Webb is the canonical instance — is a story an editor learns not to want. Flak is also manufactured to order. The model points to the rise of institutions whose entire function is to produce it: Accuracy in Media, founded by Reed Irvine in 1969, and the better-funded think tanks and monitoring operations that followed, exist to flood any outlet that strays with letters, complaints, advertiser warnings, and accusations of bias. The flak machine need not win the argument on the merits. It need only raise the anticipated cost of a deviation high enough that the editor, calculating in advance, quietly declines to pay it.
The fifth filter Herman and Chomsky originally named anti-communism: the national quasi-religion of the Cold War, an ideology so total that to be accused of softness toward it ended careers, and that supplied a ready frame for sorting the world's events into the deeds of "our side" and "theirs." In the 2002 edition's introduction, written after the Soviet collapse, they generalized the filter to the dominant ideology of the moment — and named its successor directly: the "war on terror," which by 2002 was performing precisely the same function, providing the unquestioned premise within which all respectable debate was required to stay. The fifth filter is the one that turns the other four into a worldview. It is not a topic. It is the air.
The model's signature move is that it makes predictions, and Herman and Chomsky structured the book to test them. The most rigorous test is the doctrine of worthy and unworthy victims. The model predicts that the press will lavish coverage, indignation, and demands for justice on victims of official enemies — "worthy" victims, whose suffering is useful — while treating the equivalent or worse victims of the United States and its clients as statistical background, "unworthy" of the same moral attention.
To test it, the authors did something unusual for a polemic: they counted. They compared the coverage of Jerzy Popiełuszko, a Polish priest murdered by communist secret police in 1984, against the coverage of religious figures murdered in U.S.-aligned Latin American states during the same period — among them Archbishop Óscar Romero, gunned down at the altar in El Salvador in 1980, and the American churchwomen raped and killed by Salvadoran forces later that year. The single worthy victim, Popiełuszko, received more coverage in the New York Times, Time, Newsweek, and CBS News than a hundred unworthy victims of U.S. client regimes combined — and his murderers were tried, while the Salvadoran killers operated with American aid flowing uninterrupted. The disparity was not in the facts available; it was in which facts the filters let through. The model does not claim the press lies about worthy victims. It claims the press is sincere about them, and that the sincerity is selective in exactly the direction the institutional structure predicts.
The same machinery sorts elections. Herman and Chomsky compared the U.S. press treatment of the 1982 election in El Salvador and the 1984 election in Guatemala — held under U.S.-backed governments whose armies were slaughtering civilians and whose opposition candidates were being murdered or driven into exile — with treatment of Nicaragua's 1984 election, won by the Sandinistas. The first two were framed as hopeful steps toward emerging democracy; the third was dismissed as a propaganda sham, despite an opposition slate, a substantial turnout, and international observers who rated it cleaner than its U.S.-aligned neighbors. The criterion governing the coverage was not the procedural quality of the vote but whether its outcome served Washington. A "legitimizing" election validated a client; a "meaningless" election delegitimized an enemy, and the same press applied the two labels with no apparent awareness of the contradiction.
The model also predicted credulity running the other way — the press amplifying an official enemy's guilt on thin evidence. Herman and Chomsky devoted a chapter to the "Bulgarian Connection," the claim that the KGB had directed the 1981 shooting of Pope John Paul II through Bulgarian intelligence. Promoted by the journalist Claire Sterling and amplified across American media as all but established fact, the thesis rested on the shifting jailhouse testimony of the would-be assassin, the Turkish ultranationalist Mehmet Ali Ağca, and it collapsed in an Italian courtroom for want of evidence. The press had extended to a geopolitically useful accusation against the Soviet bloc a confidence it never granted to far better-documented crimes by allied states. Worthy villains, like worthy victims, receive the benefit of the frame.
The 1992 documentary Manufacturing Consent: Noam Chomsky and the Media, directed by Mark Achbar and Peter Wintonick, dramatized the East Timor comparison with a literal stack: Chomsky standing beside a towering column of newsprint devoted to Cambodia next to a thin sheaf on East Timor (approx. 1:14:00). The same film captured the model's subtler claim through the television producer Jeff Greenfield explaining, without embarrassment, why Chomsky would never be booked on Nightline: he lacks "concision" — the ability to say something "between two commercials" (approx. 0:42:00). Ideas that fit the format are the ideas that require no evidence because everyone already shares them; ideas that need explaining are structurally filtered out not by censorship but by the clock. Concision, Chomsky argued, is itself a filter — one that guarantees the unconventional thought never gets the airtime to make its case.
The model's predictions did not expire with the Cold War; the clearest twenty-first-century confirmation arrived in 2002 and 2003. As the Bush administration built its case for invading Iraq, the prestige press — the New York Times most prominently, through Judith Miller's front-page reporting — amplified official claims about weapons of mass destruction that rested on intelligence the press could not and did not independently verify, sourced overwhelmingly to the government officials and exile-group "primary definers" the administration controlled. Every filter was visible at once: corporate owners with no appetite for an antiwar posture after September 11, advertisers and audiences inside a patriotic consensus, total dependence on official sourcing, the certainty of flak for any outlet branded unpatriotic, and the "war on terror" supplying the unquestioned frame. The handful of reporters who got the story right — Knight Ridder's Warren Strobel and Jonathan Landay — worked for a chain that owned no marquee Washington paper and sat outside the consensus the filters protected. The model had predicted exactly this distribution of right and wrong answers fifteen years before the war.
The propaganda model is frequently mistaken for a conspiracy theory, and Herman and Chomsky spent decades insisting it is the opposite. A conspiracy requires conscious coordination, secret instruction, a directing hand. The model requires none. It describes a "guided market system" in which outcomes emerge from incentives the way water finds the drain — no meeting, no memo, no list of assets. This is exactly where it diverges from, and completes, Operation Mockingbird. Carl Bernstein's 1977 reporting documented the covert version: four hundred journalists secretly carrying CIA assignments, stories planted and killed on direct instruction. That is conspiracy in the literal sense. The propaganda model's unsettling claim is that the covert apparatus is, for most purposes, redundant. The five filters produce the same compliant press through ownership and incentive that Mockingbird produced through recruitment — which is why the post-Church-Committee dismantling of the formal program changed the press coverage so little. The Wurlitzer, in the Operation Mockingbird node's phrase, learned to play itself.
This is also the precise sense in which the model relates to the engineering tradition. Edward Bernays & The Engineering of Consent supplied the demand side — the conscious craft of manufacturing a public's consent, sold to corporate and government clients as a service. The propaganda model supplies the supply side: it explains why the commercial press is the reliable conduit through which Bernaysian technique reaches the population, and why a captured channel is the default rather than something that must be arranged each time. The same division separates it from The Tavistock Institute, whose critics charge it with theorizing mass-behavioral control in the abstract; the propaganda model names the unglamorous institutional plumbing — the ad buy, the beat reporter, the proprietor's bottom line — through which any such control would actually have to operate to reach a real audience. And both descend, in Herman and Chomsky's own acknowledgment, from an older European insight: the The Frankfurt School & The Culture Industry account of the "culture industry," in which Theodor Adorno and Max Horkheimer argued in 1944 that mass-produced culture in a capitalist economy administers consciousness into conformity. What the Frankfurt theorists described philosophically, Herman and Chomsky rebuilt as a falsifiable political-economic machine.
The model was written for an era of three networks and a morning paper, and its critics' first instinct was that the internet would destroy it — that infinite channels and zero-cost publishing would dissolve the chokepoints the five filters describe. The opposite happened. Ownership reconcentrated, this time around a handful of platforms whose scale exceeds anything Henry Luce commanded. Advertising did not loosen its grip; it became the entire business model, refined into surveillance-driven attention markets in which the product sold to advertisers is no longer the affluent reader but the precise behavioral profile. The result is what the The Dead Internet Theory thesis describes as a sixth filter: algorithmic curation and bot amplification performing, at machine speed and without human deliberation, the boundary-setting that editors and advertisers once did by hand.
The recommendation engine is the third and fourth filters fused and automated. It decides which sources are visible and which are throttled — sourcing — and it learns, from engagement signals and advertiser-safety rules, which content provokes costly flak and demotes it before any human complains. Where the old model required a publisher to internalize the limits of acceptable discourse, the new one encodes those limits in a ranking function nobody can read. The "manufacture of consent" that Lippmann imagined as the deliberate work of a specialized class, and that Herman and Chomsky relocated to the structural incentives of corporate media, has in the algorithmic age been delegated to systems that manufacture consensus as a byproduct of optimizing for time-on-site — and that can flood the channel with synthetic participants until the boundary of debate is set not by what people believe but by what the bots and the feed make visible.
The propaganda model has serious critics, and their case is not weak. The most damaging charge is that the model is functionalist in the pejorative sense: it explains every media outcome by the function that outcome serves for power, which makes it nearly unfalsifiable. Adversarial coverage proves the system's legitimating need to appear free; conformist coverage proves the filters; either way the model wins. The sociologist Daniel Hallin and others argued that the press's relationship to power is genuinely more contested than a filter model allows — that there is a "sphere of legitimate controversy" inside which real adversarial journalism happens, and that the model's machinery erases the agency of working reporters who break stories at real personal cost. The empirical objections are concrete: Watergate brought down a president; the Pentagon Papers were published over the government's furious objection; Seymour Hersh exposed My Lai and Abu Ghraib; the Times and Post published Snowden's NSA disclosures; the Panama and Pentagon leaks embarrassed the powerful worldwide. A press that is a mere transmission belt for elite consensus should not, the critics argue, keep producing the journalism that humiliates elites. And the model's examples, they add, are cherry-picked — a handful of paired cases chosen because they fit, with the contrary cases left out.
Herman and Chomsky's reply is that the critics have mistaken what the model predicts. It does not predict that every story serves power or that adversarial reporting never happens. It predicts the boundaries of debate — the range of premises that remain unquestioned across the whole spectrum of respectable opinion, including its adversarial wing. Watergate, on this reading, is not a refutation but a confirmation: the press savaged Nixon for spying on the powerful Democratic Party, while the same years' far graver crimes against the powerless — the FBI's COINTELPRO campaign that helped get Black activists killed — drew almost no comparable outrage, because the victims were unworthy. The Vietnam debate raged endlessly over whether the war was winnable or too costly, and almost never over whether it was criminal aggression — the premise the filters quietly held in place. Adversarial reporting on tactics and efficacy is not only permitted but useful; it demonstrates the system's freedom while leaving the foundational assumptions untouched. As for functionalism, Herman in particular insisted the model is testable and was tested — the worthy/unworthy victim comparisons are quantitative predictions about differential coverage that could have come out the other way and did not. The model's claim is statistical and structural, not a thesis that any single article is propaganda. It is a prediction about the shape of the whole, and the shape, its authors maintained, keeps matching.
The deepest form of the reply is the one Chomsky gave the BBC's Andrew Marr in 1996, after Marr objected that he, personally, was conscious of no self-censorship. "I'm not saying you're self-censoring," Chomsky answered. "I'm sure you believe everything you're saying. But what I'm saying is, if you believed something different, you wouldn't be sitting where you're sitting." This is the model's answer to the agency objection compressed into a sentence. The filters do not work by overriding the consciences of honest journalists; they work by selecting, across a career, for the journalists whose sincere convictions already fit — the columnist who rises is the one who never had to be told what not to write. Agency is real. It is precisely the thing the structure has already shaped, long before any individual story is assigned. The objection that the model erases the reporter's autonomy misses that the model is a claim about which autonomous reporters get hired, promoted, and read.
What neither side disputes is the stakes. If Lippmann was right that consent must be manufactured for a mass democracy to function, and Herman and Chomsky were right about who does the manufacturing and why, then the central institution of self-government is structurally configured to produce a public that freely agrees to what it was never permitted to seriously question. The disagreement is only over how tight the configuration is — and whether the journalist standing inside it is an agent of conscience or a component of the machine. The model says: usually both, and the machine usually wins the cases that matter.