The Persuasion Machine: 100 Years of Marketing Tricks, Who Runs Them Best, and How to Stop Falling for Them

By Milad Ghobadibeygvand, BScN (Western University, 2014) · Published July 26, 2026 · Zeus eBikes Canada

You were marketed to before you finished breakfast today, and the people who did it did not leave their methods to chance. For a hundred years the persuasion industry has tested, measured, refined and occasionally confessed — in courtrooms, in leaked memos, in refund ledgers — exactly how it moves you. This article is not their account of themselves. Marketers' descriptions of their own tricks are, after all, marketing. This is the other file: what the receipts prove.

The receipts have been arriving fast. In September 2025, Amazon paid $2.5 billion — the largest settlement in the US Federal Trade Commission's history — over a Prime cancellation maze its own designers named the "Iliad Flow": four pages, six clicks, fifteen options between you and leaving (FTC settlement, September 2025). In January 2026, Canada's Federal Court of Appeal upheld a $38.9-million penalty against Cineplex over a $1.50 booking fee that appeared only after the advertised price (Competition Tribunal, 2024; Federal Court of Appeal, January 21, 2026). The machine is a century old. The bills are new.

What follows reverse-engineers the whole apparatus: where the techniques came from, which ones actually survive scientific testing, how every segment of your family is targeted differently, what measurably happens when a country bans a trick, how the same machinery runs politics — including the world's first news-banned election, which was ours — and the defence kit with actual evidence behind it. No products are recommended anywhere in this piece; this is public-interest work. Our only stake is disclosed at the end, where we show you our own hand.

---

The Week the Machine Got Billed

For most of a century, the persuasion industry's tricks cost it nothing; in the past three years, courts and regulators began pricing them, and the price list is the best map of the toolkit we have ever had. The FTC's $2.5-billion Amazon settlement (September 2025), Epic Games' $520-million penalty (2022), Cineplex's $38.9-million drip-pricing judgment upheld on appeal (January 2026): read the enforcement record and you are reading a confession, in dollars, of which techniques were worth building.

Look at what each bill itemizes. Amazon's "Iliad Flow" was not an accident of bad design — internal documents surfaced in the case show the company knew the cancellation maze was holding subscribers in, and named it after an epic about a ten-year war (FTC v. Amazon, 2025). Epic's fine covered interface designs that steered children into purchases they did not intend (FTC, December 2022). Cineplex's penalty was calibrated by Canada's Competition Tribunal to approximately the total the undisclosed $1.50 fee had collected — a rare, exact public accounting of what one small trick earns: $38.9 million from a buck fifty (Competition Tribunal, September 2024; Federal Court of Appeal, January 21, 2026). And the European Union's consumer-protection network formally notified Temu in November 2024 that its platform ran fake discounts, "pressure selling," and "forced gamification" — a regulator's plain-language inventory of a modern storefront's persuasion stack (European Commission CPC Network, November 2024).

Hold that enforcement record beside a number from the other side of the ledger: the world now spends more than US$1 trillion a year on advertising — for the first time in history — with the top fifty advertisers alone spending $291 billion, Amazon at the head of the table, and one company, Alphabet, collecting over $200 billion of it (eMarketer, 2025; Ad Age Datacenter, 2024). A trillion dollars is not spent annually on something that doesn't work. The rest of this article is about what, precisely, that money buys — established the hard way, from outcomes.

---

1917–1929: The Invention — War Propaganda Becomes Marketing

Modern marketing was not invented to sell soap; it was invented to sell a war, and then the men who built it took the machine private. In April 1917, the United States created the first industrial-scale propaganda agency — the Committee on Public Information under George Creel — and within eighteen months it had organized roughly 75,000 volunteer speakers, the "Four Minute Men," who delivered more than 7.5 million four-minute speeches in cinemas, churches and union halls, reaching a cumulative audience counted in the hundreds of millions (Library of Congress; CPI records). Posters, films, press placement, testimonial, fear appeal, social pressure: the full modern toolkit, assembled under government contract, aimed at a democracy's own citizens.

A Four Minute Man addresses a cinema audience, 1917 (recreation). The Committee on Public Information fielded roughly 75,000 volunteer speakers who delivered more than 7.5 million short speeches (Library of Congress).

Among the committee's alumni was a young man named Edward Bernays — Sigmund Freud's nephew — who drew the century's most consequential commercial conclusion: if these methods could sell a war, they could sell anything. His demonstration case still astonishes. In 1929, hired by the American Tobacco Company to break the taboo against women smoking in public, Bernays staged a spectacle: at the Easter parade on Fifth Avenue, March 31, 1929, young women — including his own secretary, Bertha Hunt — stepped out and theatrically lit Lucky Strikes, which the press was primed to describe as "torches of freedom." The New York Times ran the story the next morning: "Group of Girls Puff at Cigarettes as a Gesture of 'Freedom.'" Neither Bernays nor his client appeared in any coverage (History Today; press archive). He had not bought an advertisement. He had manufactured an event, attached a product to a liberation movement, and let journalism distribute it as news.

Easter Sunday, 1929 (recreation): the "Torches of Freedom" stunt was staged for the press by Edward Bernays on American Tobacco's account — and reported the next morning as news, not advertising.

Now the result, measured the way this article measures everything: women's share of cigarette consumption, about 5 percent in 1923, reached 12 percent by 1929 — and kept climbing for four decades (industry consumption data via tobacco-history archives). Bernays wrote the method down in a book he titled, without embarrassment, Propaganda (1928), and later coined the phrase "the engineering of consent." Meanwhile a quieter revolution was running in parallel: Claude Hopkins' Scientific Advertising (1923) preached coupons and tracked responses — every ad a measured experiment — which makes Hopkins the great-grandfather of the A/B test and one of the few early admen this article can cite, because his method produced receipts rather than boasts.

One more foundation stone, darker than the rest. When science linked smoking to cancer, the industry that Bernays helped build did not argue the science — it industrialized doubt about it. A 1969 Brown & Williamson planning document, pried loose decades later among more than 14 million pages of internal tobacco files now archived at UCSF, states the strategy in one sentence: "Doubt is our product, since it is the best means of competing with the 'body of fact' that exists in the mind of the general public." That memo is the Rosetta Stone of a technique this article will meet again in politics: when you cannot win the argument, market uncertainty about the argument.

---

The Founding Hoax: Subliminal Advertising Never Worked

The most famous marketing trick of the twentieth century — hidden frames flashing "EAT POPCORN" past your conscious mind — is a confessed fabrication, and the confession is on the record. In 1957, market researcher James Vicary announced that imperceptible messages inserted into a New Jersey cinema's films had raised popcorn sales 57.5 percent and Coca-Cola sales 18.1 percent. Panic followed; legislatures moved; the phrase "subliminal advertising" entered every language. In 1962, Vicary told Advertising Age the truth: the experiment was a "gimmick" concocted to rescue his failing consultancy, with data too small to mean anything. Independent replications found no meaningful effect then, and have found none since (Advertising Age interview, 1962; replication literature).

The scene of the crime that never happened (recreation): James Vicary's 1957 subliminal "experiment" panicked legislatures worldwide — and in 1962 he told Advertising Age it was a "gimmick" invented to save his failing firm.

The aftermath is the instructive part, because the myth outlived its own confession by seventy years and counting. Canada's broadcast law still prohibits subliminal devices (Television Broadcasting Regulations, 1987) — a ban on a technique that never worked — while the techniques that demonstrably do work were never hidden at all. That is this article's first defensive lesson, and it comes free: the persuasion you should fear is not secret frames; it is the visible architecture — the anchor price, the prechecked box, the countdown clock — operating in plain sight precisely because plain sight is where attention doesn't look. Whenever a marketing scare sounds like mind control, check for a Vicary; whenever a technique sounds too mundane to matter, check for a default.

---

The Golden Costume: What the Mad Men Actually Sold (1929–1972)

Between the wars' propaganda and the laboratory era sits advertising's mythologized golden age — and read by results instead of memoirs, its true product was not slogans but delivery systems and identity engineering. The receipts are cultural, and they are enormous.

Start with the delivery system so successful it named a genre. In 1933, Procter & Gamble built a radio drama, Ma Perkins, for the explicit purpose of selling Oxydol soap flakes to the women listening at home — and the format it pioneered became so identified with its sponsors' product category that the world still calls the genre the soap opera. That is the machine's signature move stated plainly: when the audience won't come to the ad, build the entertainment around the ad, and make the container so good nobody leaves. Every "free" app on your phone is Ma Perkins with a telemetry layer.

Then the identity engineering — and the century's most audited case is a costume change in the most literal sense. Philip Morris launched Marlboro in 1924 as a women's cigarette, slogan "Mild as May," and by mid-century the brand was a minor player with a problem: the new filtered cigarettes read as feminine to the male smokers worried (correctly) about their health. Leo Burnett's answer, first run in 1954–55, was to bolt the most masculine image in the American catalogue — a working cowboy — onto the identical product. No health claims, no filter science: Burnett deliberately refused to argue about safety at all, because arguing acknowledges the fear. The measured result: within roughly a year the brand had vaulted from obscurity into the top ranks of American cigarette sales, on its way to becoming the best-selling cigarette on Earth — a trajectory documented across the industry histories and the tobacco-litigation record. Nothing about the cigarette changed. The buyer's reflection in it changed. Hold that beside the LEGO file in the segment chapter below: re-gendering a product is a routine industrial operation, run in both directions, whenever the ledger recommends it.

The wordless sell (recreation): the cowboy re-gendered a 1924 women's cigarette without one health claim — within about a year of the 1954–55 campaign the brand had vaulted from obscurity into America's best-selling ranks.

The era's own theorists get the Vicary treatment here — audited, not admired. Rosser Reeves preached the Unique Selling Proposition and hammered one claim per product ("fast, fast, fast relief"); he also claimed his campaigns multiplied sales, which is practitioner testimony we cannot audit, so by this article's rules it stays in the intent column. Vance Packard's 1957 bestseller The Hidden Persuaders convinced a generation that depth psychologists were steering them through their unconscious — but the motivational-research industry it feared produced no replicable audited wins, while the visible levers (anchors, defaults, repetition, distribution) kept compounding; the book's real measured effect was on the public, priming it to believe Vicary's hoax months later. And the age's genuinely deserved monument is the counterintuitive one: Doyle Dane Bernbach's 1959 Volkswagen campaign — "Think Small," a tiny German car photographed in an ocean of white space, copy that opened by conceding the car was ugly and slow — which the industry's own trade bible, Advertising Age, ranked the #1 campaign of the twentieth century in its 1999 retrospective. Its mechanism is the one this article's defence kit weaponizes in reverse: volunteered weakness buys belief for every claim that follows. Honesty, it turns out, is a technique too — the only one that survives full information, which is why it is the one this publication bet its own marketing on.

The era closed by inventing the word for what came next. In 1972, Al Ries and Jack Trout published the "Positioning" series in Advertising Age, formalizing the idea that the battle is not for shelf space but for a slot in the prospect's mind — own one word, or own nothing. Positioning is segmentation's twin, and together they hand the microphone to the machine's next chapter: once you accept that the product lives in the mind, the logical next step is to study whose mind, which is where the segment files begin.

---

The Levers That Survive Testing — and the Ones That Didn't

Strip away a century of jargon and the machine runs on about a dozen levers, and the honest scorecard — behavioural results only, marketer testimony excluded — sorts them into three bins: proven by measured behaviour, contested under replication, and folklore. This section is the article's engine room, and it includes the part the marketing industry never volunteers: some of its own favourite science collapsed when independent researchers reran it.

The anchor at work: the struck-through price does the selling, not the product. Canada's Competition Act now requires "was" prices to be genuine — Canadian Tire paid $1.29 million in Quebec over inflated regular prices.

Proven by results

• Defaults — the strongest lever ever measured. The proof is a natural experiment nobody designed: European countries with opt-in organ donation (you must sign up) versus opt-out neighbours (you are enrolled unless you object). Germany, opt-in: consent around 12 percent. Austria — similar culture, similar wealth, different default: over 99 percent. When researchers Eric Johnson and Daniel Goldstein assembled the comparison in Science ("Do Defaults Save Lives?", 2003), online experiments confirmed it: the pre-set choice roughly doubles agreement. Nobody was persuaded of anything. The box was just already checked. Every subscription trial, every prechecked add-on, every auto-renewal is this lever, monetized.
• Anchoring — the first number wins. In Tversky and Kahneman's classic demonstration (Science, 1974), spinning a rigged wheel of fortune before an unrelated estimate dragged people's answers toward the meaningless number they had just seen (a "10" produced average guesses of 25 percent; a "65" produced 45 percent). Retail runs this daily as the "was $199" strike-through: the anchor is doing the selling, not the product. Our own field audit, below, found was-price anchors in the raw HTML of Canadian sale pages by the dozen.
• The decoy — a third option that exists to lose. Huber, Payne and Puto showed in 1982 that adding an option nobody would choose (worse than one alternative in every way) reliably shifts buyers toward the "dominating" choice — the mechanism behind the medium popcorn priced a quarter below the large.
• Social proof — measured on hotel towels. When a hotel sign said most previous guests reused towels, reuse rose 26 percent versus the standard eco-appeal; saying most guests in this exact room had done it pushed the lift to 33 percent (Goldstein, Cialdini & Griskevicius, Journal of Consumer Research, 2008 — a field experiment with laundered towels as the endpoint, which is why it appears here while most "influence" literature does not). "37 people are viewing this hotel" is the same lever with worse manners.
• Scarcity — the cookie-jar result. Identical cookies were rated more desirable when the jar held two rather than ten (Worchel, Lee & Adewole, 1975). Every countdown timer and "only 3 left" badge rents this finding.
• Loss framing. Kahneman and Tversky's prospect theory (1979) established that losses loom roughly twice as large as gains — which is why insurance, extended warranties, and political attack ads all sell the loss, not the gain.
• Variable rewards — the slot-machine schedule. B.F. Skinner established that unpredictable rewards produce the most persistent behaviour ever conditioned; loot boxes run precisely this schedule, spending on them correlates meta-analytically with problem gambling, and adolescents are the most exposed cohort. This lever's regulation record — Belgium's failed ban — appears in the Atlas below.
• Commitment — the foot in the door. Homeowners who first agreed to display a tiny sign became dramatically more likely to accept a huge ugly one two weeks later (Freedman & Fraser, JPSP, 1966). Free trials, quizzes, and "just create an account" flows are industrialized versions of the small first yes.

Contested — and the industry's own replication scandal

Honesty requires the other bin. "Priming" — the claim that incidental words or images unconsciously steer behaviour, a staple of marketing conference keynotes — failed catastrophically under replication: a 2017 forensic analysis of the priming chapter in the field's most famous book found the underlying studies statistically incapable of supporting their conclusions, and Daniel Kahneman himself replied publicly, with a scientist's grace: he had "placed too much faith in underpowered studies," having already urged colleagues in a 2012 open letter to rebuild the field with a "daisy chain" of replications (Schimmack, Heene & Kesavan, 2017; Kahneman correspondence). The "nudge" industry met the same audit: a 2022 meta-analysis in PNAS celebrated choice architecture's average effects, and a response in the same journal found that after correcting for publication bias, the remaining evidence for nudging-in-general was indistinguishable from zero (Mertens et al., 2022; Maier et al., 2022). The honest synthesis: the closer a lever sits to changing the choice environment itself — defaults, prices, fees, friction — the better it survives audit; the closer it sits to whispering at your unconscious, the worse. Manipulation is real. It is just far more architectural, and far less mystical, than either its sellers or its scaremongers claim.

---

The Segment Files: Boys, Girls, Teens, Seniors, and Your Postal Code

The machine does not address "consumers"; it addresses segments — and the deepest proof is a study where researchers ran identical ads with no targeting at all and watched the platform segment humanity on its own. In 2019, a Northeastern-led team bought real Facebook ads for housing and jobs, aimed at broad audiences, and observed delivery: the platform's optimizer skewed who saw them by race and by gender anyway — lumber jobs routed to white men, secretarial jobs to Black women, housing ads split along racial lines — because the delivery algorithm had learned which segments "engage" (Ali et al., 2019). Three years later that mechanism became the first algorithm the US Justice Department ever forced a company to rebuild under the Fair Housing Act (DOJ–Meta settlement, June 2022). Segmentation is not a menu advertisers choose from. It is the machine's native language.

Boys and girls: the aisle was a business decision

Anyone who believes the pink aisle is natural should study the 1970s, when it briefly wasn't. Second-wave feminism pushed retailers to de-gender: the Sears catalogue consciously showed girls with construction sets and boys at play kitchens, and LEGO's now-famous 1981 ad presented a small girl in jeans and sneakers holding her own creation over the caption "What it is is beautiful" — no pink, no princess, no gender in the copy at all. Then the industry reversed itself. Through the 1980s and 1990s, children's marketing re-segmented harder than ever — princess lines, action lines, colour-coded everything — for a reason no one hid: two gendered aisles sell more than one shared one (hand-me-downs die when the sibling is the "wrong" gender). Content analyses of the resulting decades find boys' marketing scripting skill, heroism and expertise while girls' scripts appearance, caretaking and domesticity — a curriculum in disguise, renewed with every toy catalogue (research reviews summarized in The Conversation, 2021; LEGO itself pledged in 2021 to strip gender bias from its marketing).

Two markets, one child (staged illustration). The industry de-gendered in the 1970s, then deliberately re-segmented in the 1980s–90s: two gendered aisles sell more than one shared one.

The gendered machine follows its subjects into adulthood and bills them. New York City's consumer-affairs department audited roughly 800 near-identical products in 2015 (From Cradle to Cane): the women's version cost more 42 percent of the time, against 18 percent for men's, averaging 7 percent more overall and 13 percent more for personal care — the "pink tax," measured on shelves, not asserted (NYC DCA, 2015). Same razor, same cart, different target file, different price.

Teens and young men: the highest-value nervous systems

Adolescents combine maximum reward-sensitivity with minimum consumer defence, and the machine has always known it. The modern files: loot boxes placing the slot-machine schedule inside children's games (see the Atlas for Belgium's failed ban); social feeds engineered for the adolescent dopamine cycle — a file this publication documented separately in our teen-loneliness investigation and the parents' guide; and the newest chapter, sports betting, which arrived in Ontario with 2022's legal igaming market and saturated broadcasts so thoroughly that the province's regulator intervened — from February 28, 2024, athletes and celebrities appealing to minors are banned from Ontario igaming advertising (AGCO Registrar's Standards). A rule aimed at one segment is the regulator telling you, in writing, which segment the ads were built to reach.

Seniors: the trust segment

At the other end of life, the machine hunts accumulated trust and diminished processing speed. The cleanest documented case sits in politics rather than commerce: the WinRed fundraising platform's prechecked recurring-donation boxes — buried under paragraphs of urgent copy — extracted repeat withdrawals from donors who believed they had given once, with the harm concentrated among retirees on fixed incomes, veterans, and at least one donor in hospice. The observed result: $122.7 million refunded, more than one dollar in ten raised, and a unanimous 6–0 Federal Election Commission recommendation to ban the prechecked box (NYT investigation, 2021; FEC). The same interface trick Amazon ran on Prime subscribers, aimed at grandmothers' bank accounts, wearing a flag.

Your postal code, your income, your grief

Geodemographic systems have sorted households since the 1970s — Canadian marketers can buy their way into segment taxonomies that classify every postal walk by income, ethnicity, and "lifestyle cluster" — and digital delivery sharpened the blade to individual level: the Ali study above shows the sorting now happens even when no one asks for it. The uncomfortable summary of this entire section: you are not marketed to as a person; you are marketed to as a file — and the file knows whether you are a boy being sold heroism, a girl being sold beauty, a teen being sold odds, a senior being sold urgency, or a postal code being sold what postal codes like yours already bought.

The country file: when the machine chose rural

Most claims about urban-versus-rural targeting are unaudited folklore — trucks sold on country radio, condos sold on transit screens — and this article's method obliges us to say so. The audited case is darker, and it is the most consequential segment file in the modern record. Purdue Pharma's OxyContin campaign concentrated its sales force on the highest-volume prescribers, and the geography where the resulting death curve bent first was rural — the mining and manual-labour counties of Appalachia lead the overdose record. The receipts are judicial, twice over: a Purdue affiliate pleaded guilty in 2007 to misbranding OxyContin from 1996 through 2001 as less addictive than other opioids, the company pleaded guilty again in 2020, and in December 2024 the US Justice Department announced a $650-million resolution with McKinsey & Company — 75 engagements with Purdue, including 2013 advice on how to "turbocharge" OxyContin sales — with a former senior partner charged with obstruction for deleting records (DOJ, 2024). A segment file is a neutral instrument the way a scalpel is: the same precision that sells step-through bikes to seniors sold a time-release opioid to the counties least equipped to absorb it. That is why this chapter exists.

---

The Digital Re-Costume: Dark Patterns, Drip Pricing, and Our Field Audit

Nothing fundamental changed when marketing went digital — the levers are the 1950s set — but three things multiplied: the testing speed (millions of variants), the precision (the segment files above), and the interface's power to make the lever physically hard to escape. Regulators coined a name for that last one — dark patterns — and the recent case law reads like a user manual written in fines.

The cancellation maze. Amazon's internal "Iliad Flow" required four pages, six clicks and fifteen options to leave Prime while sign-up took one; the FTC's $2.5-billion settlement (September 2025) included $1.5 billion in direct refunds to subscribers who were enrolled or retained against their intent. The child-steering interface. Epic's $520-million penalty (2022) covered designs that converted children's button presses into purchases. The dripped fee. Cineplex's $1.50 online booking fee — revealed after the advertised ticket price — became Canada's defining case: the Competition Tribunal set the penalty at roughly the fee's entire take, $38.9 million, and the Federal Court of Appeal affirmed in January 2026. Since the 2022–2024 Competition Act amendments, advertising a price no one can actually pay (unless the addition is a government tax) is explicitly deceptive in Canada, discount claims must be genuine, and environmental claims now carry a reversed onus — the business must hold the proof before it speaks (Bill C-59, June 2024). The pressure platform. The EU's consumer-protection network formally cited Temu (November 2024) for fake discounts, pressure selling and "forced gamification" — spin-the-wheel mechanics you must play through to shop. The undisclosed endorsement. Canada's Ad Standards updated its influencer-disclosure guidelines in October 2025 (now covering AI-generated influencers and child audiences); the Competition Act's material-connection rules apply, and US enforcement runs to ~$53,000 per violation. The influencer era's defining trick is manufactured intimacy — an ad wearing a friendship's face — and the disclosure hashtag is the entire legal boundary between the two.

Our field audit: what the machinery looks like from a Canadian browser

Following this article's method, we ran our own small observation on July 26, 2026: a single automated fetch of 12 Canadian-facing storefront pages (major domestic retailers' deal pages, two global fast-fashion platforms, ticketing and travel sites), counting urgency machinery visible in the raw served HTML — countdown/timer markers, scarcity phrases ("only X left"), urgency copy ("ends tonight"), was-price anchors, and social-proof nudges ("people are viewing"). Method notes, honestly stated: one fetch per site, static HTML only, so anything rendered by JavaScript is invisible to us — every count is a floor — and pattern-matching counts machinery present in the page source, not claims we individually verified on screen. Raw data is archived and the method is fully replicable by any reader with a terminal.

The floor was busy. Nine of twelve sites answered our fetch with HTML, and among those nine, six carried countdown/timer machinery, five carried urgency copy, and three carried was-price anchors in the first served page alone — one home-goods giant's daily-sale page held 37 was-price anchors, 30 timer markers and 13 social-proof nudges in its raw source. Just as telling was who we couldn't audit: three major Canadian retailers refused the request outright, and Temu — the platform EU regulators cited for pressure selling — served a 2-kilobyte JavaScript shell containing nothing at all. The most aggressive persuasion architectures on the consumer internet are structurally invisible to simple public audit, which is precisely why European enforcement had to inspect them with full browsers and legal powers. In an industry this measurable, opacity is a choice.

12:47 AM, engineered urgency (staged illustration). Our July 2026 field audit found countdown machinery in the served HTML of six of the nine Canadian-facing storefronts that answered a simple public fetch.

---

Who Is Best at It

Rank practitioners the way this article ranks everything — by measured results, revealed methods, and paid penalties — and a short list emerges that has little to do with creative awards. "Best" here is a warning label, not a compliment.

• The platform duopoly-plus, for building the auction. Alphabet is the first company in history to collect over $200 billion a year in advertising; Meta's feed machinery — documented from its own files in our platform investigation — turned attention itself into inventory. They are not the best at making ads. They are the best at owning the room every ad must rent.
• Amazon, for weaponized friction. The world's largest advertiser is also the firm that named its cancellation maze after the Iliad and paid the largest dark-pattern settlement ever. Mastery of the default and the maze, priced by a court at $2.5 billion.
• Tobacco, for the doubt patent. The industry that wrote "Doubt is our product" ran the most documented persuasion operation in corporate history — 14 million pages of receipts — and its playbook (fund uncertainty, demand "balance," delay forever) has been licensed by every industry facing inconvenient science since; that lineage is documented at book length in the historians' work on manufactured doubt and appears throughout our 100 Lies ledger.
• The gambling-mechanics industry, for conditioning children. Loot boxes put Skinner's variable-ratio schedule inside games rated for twelve-year-olds, survived Belgium's ban at an observed 82 percent non-compliance rate, and migrated to battle passes when pressed. No sector has monetized a reinforcement schedule more efficiently.
• Pharma, for the two-country trick. Direct-to-consumer prescription-drug advertising — "ask your doctor about" — is lawful essentially only in the United States and New Zealand; Canada prohibits full product-claim ads. The industry's genius was making the world's most restricted product feel like a consumer good in the two markets that allowed it; every American drug ad leaking into Canadian feeds is a free demonstration that the same molecule can be marketed like a phone plan where law permits.
• De Beers and the heritage masters, for inventing demand itself. The diamond engagement ring as "necessity," the Marlboro cowboy that re-gendered a filtered cigarette, Red Bull's transformation of a beverage company into a media empire — the individual case files, with their documented deceptions where they exist, live in the 100 Lies archive; they earn their place here as proof that the highest form of the craft sells a feeling that outlives every fact-check.

---

The Rulebook Atlas: When Countries Ban a Trick, What Actually Changes

Here is the section this article's method was built for. Every advertising rule is a natural experiment: a jurisdiction switches one lever off, the rest of the world doesn't, and the difference — measured in spending, consumption, prevalence — is the closest thing persuasion science has to a controlled trial at national scale. The atlas below reports only rules whose results were measured, including the failures, because the failures teach the most.

Read the atlas twice and two patterns surface. First: switching a lever off changes behaviour, measurably, at national scale — Quebec's children ate millions fewer fast-food meals, Australia's smokers quit measurably faster, Chile's families bought a quarter less sugar-water. Advertising's defenders sometimes claim it only shuffles brand share without changing category demand; the atlas is the refutation, delivered by the industry's own fiercest lobbying against every one of these laws. Nobody spends millions fighting a rule that wouldn't change anything. Second: rules bind exactly as far as their enforcement and their system reach. Chile changed purchases but not obesity because the food system reformulated around the rule. Belgium changed nothing because nobody enforced it. China's superlative ban moved claims sideways into implication. The lesson for Canada's young enforcement era is written in those three rows.

And ask the atlas the plainest version of the question: what happens to a society where a detergent cannot legally claim it washes whitest? In China that is not a thought experiment but statute: the claim is banned even if provable, and commerce continued; the persuasion simply migrated from words to pictures, from claims to celebrities. Persuasion behaves like water. Regulation doesn't drain it; it redirects the flow — which is why the defence kit two sections down matters more than any single rule.

---

The Political Wing: From the Daisy Ad to the News-Banned Election

Everything in this article sells candidates and policies exactly the way it sells detergent — same levers, same testing, same segment files — with one difference that should keep you up at night: in commerce the worst outcome is a bad purchase, and in politics it is a government. The political record, reverse-engineered from results, splits into what the machine measurably does, what it does not do, and the newest Canadian chapter, which happened to be a world first.

The canon, measured where possible

The "Daisy" ad — a child counting flower petals dissolving into a nuclear countdown — aired exactly once, on September 7, 1964, and then travelled the country for free as news coverage, reaching an estimated hundred million people; it never named Barry Goldwater, and it didn't need to (Britannica; Smithsonian). It taught politics the Bernays lesson: manufacture the event, let journalism distribute it. Twenty-four years later the Willie Horton campaign taught the darker one — fear plus race, laundered through an "independent" committee — and its architect, Lee Atwater, dying of cancer in 1991, apologized in Life magazine for its "naked cruelty." The confession outlived the election it won. And the vocabulary engineers proved a phrase can outlast its own author's conscience: Frank Luntz rebranded the estate tax as the "death tax" and advised in a 2002 memo that "climate change" sounded less frightening than "global warming" while urging emphasis on scientific uncertainty — tobacco's doubt patent, relicensed — before telling a Senate committee in 2019, in his own words: "I was wrong." The phrases he built are still in circulation. Retractions don't trend.

September 1964 (recreation): the "Daisy" ad aired exactly once, then travelled the country for free as news coverage. It never named the opponent. It didn't need to.

Canada owns the cleanest paired experiment in negative advertising anywhere. October 1993: the Progressive Conservatives aired an ad built around unflattering close-ups of Jean Chrétien's face, partially paralyzed since childhood by Bell's palsy, over the question "Is this a Prime Minister?" The country recoiled; the ad was pulled within a day; Chrétien's answer was devastating — "God gave me a physical defect... I've accepted that since I was a kid" and, in its most-quoted form, that he had half a face but a whole brain — and the governing party fell from 169 seats to two. Eighteen years later, the same party's successors ran "Just Visiting" and "He Didn't Come Back For You" against Michael Ignatieff — sustained, disciplined, cruel-but-deniable definition of an opponent — and it worked completely. The pair is the whole science in miniature: attack ads are neither magic nor poison; they are a lever that detonates on the sender when it violates the audience's decency floor, and defines the target when it doesn't.

What the machine measurably does now — and doesn't

The modern political machine's documented outputs are testing volume and money, not mind control. Obama's 2012 operation ran email like a laboratory — 166 audience segments tested in a single October day — and its best-tested subject line, the shruggingly casual "I will be outspent," raised roughly $2.6 million on its own, inside an online program that raised over $500 million (campaign testing records reported post-election). Trump's 2016 digital operation ran 5.9 million ad variants against Clinton's 66,000 — its own director's count, consistent with platform reporting — killing losers and feeding winners hourly. And the fundraising interface learned commerce's worst trick: WinRed's prechecked recurring-donation boxes produced $122.7 million in refunds and a unanimous FEC rebuke. Those numbers — segments, variants, refunds — are the machine's real fingerprints.

Against them, print the deflation column, because the no-cherry-picking rule applies to fear too. When the UK Information Commissioner finished the largest data-politics investigation ever conducted — servers seized, three years of forensics — her conclusion on Cambridge Analytica was that the firm's fabled psychographic weapons were "exaggerated" hype and that it did not work on Brexit beyond initial inquiries (ICO, October 2020). Randomized experiments agree with the forensics: across 59 trials of real presidential ads on 34,000 people, persuasion effects were small "regardless of context, message, sender, or receiver" (Coppock, Hill & Vavreck, Science Advances, 2020), and a two-million-person field experiment on the 2020 US campaign moved turnout by approximately nothing (Aggarwal et al., 2023). The honest synthesis mirrors the commercial one exactly: single political ads barely persuade; the architecture wins elections — who is defined early (Ignatieff), what is on the agenda at all (the media-salience mechanism established by McCombs and Shaw in 1972), which frame owns the words (Luntz's vocabulary), what the defaults are (the prechecked box), and what the window of the sayable contains (the Overton window, named for the analyst who described how positions move from unthinkable to policy). Architecture, not whispering. Again.

One political technique has receipts so clean it deserves its own paragraph: manufactured grassroots — social proof, the towel-experiment lever, aimed at regulators instead of shoppers. When the FCC took public comments on repealing net neutrality in 2017, it received 22 million; the New York Attorney General's forensic investigation later established that nearly 18 million were fake, that an industry coalition, Broadband for America, had paid about $4.2 million to generate 8.5 million of them using real people's identities without consent, and that the lead-generation firms involved had run the same play in more than 100 other advocacy campaigns, including proceedings at the EPA (NY AG report, May 2021; $4.4-million settlements with three firms followed). The word for this is "astroturfing" — fake grass roots — and its audit matters because it inverts the usual direction of manipulation: here the machine impersonated the public to the government, using the same social-proof mechanics that "37 people are viewing this" uses on you. When you weigh "thousands of comments demand X," the FCC file is the reason to ask who typed them.

The fifty-year file: when the machine was aimed at whole populations

Now the gravest drawer in the archive — the last half-century's documented cases where the identity techniques from the commercial chapters (build an in-group, brand an out-group, manufacture the event, control the agenda) were pointed at entire populations to sell them something against their own interest. This file is ranked the way everything here is ranked: by adjudication. And at the top sits the only marketing campaign ever measured in lives per broadcast tower.

Rwanda, 1994, is the definitive case — legally and scientifically. Radio-Télévision Libre des Mille Collines ran textbook identity marketing on its own audience: an in-group brand ("Hutu Power"), an out-group dehumanized as inyenzi — cockroaches — and a drumbeat of fear, testimonial and social proof. On December 3, 2003, a United Nations tribunal convicted the station's founders and an allied newspaper editor of direct and public incitement to commit genocide — the first media executives so convicted since the Nuremberg tribunal condemned the publisher of Der Stürmer in 1946 (ICTR, Nahimana et al.). Then the economists arrived with this article's exact method. Because hills block radio waves, RTLM's signal reached some villages and not their neighbours — a natural experiment drawn in topography — and when David Yanagizawa-Drott mapped reception against prosecution records, villages inside the signal showed measurably higher participation in the killing: an estimated 51,000 perpetrators, roughly 10 percent of total participation, attributable to one radio station (Quarterly Journal of Economics, 2014). No court and no journal has ever priced a marketing channel more precisely. That is what the segment-and-identity machinery does when a state gives it a monopoly and a target.

The geography of a broadcast signal (illustration). Hills blocked RTLM's signal; villages inside reception showed measurably higher participation in the 1994 genocide — roughly 51,000 perpetrators attributable to one station (Yanagizawa-Drott, QJE, 2014).

Wars get product launches. In October 1990, a tearful 15-year-old told a US congressional caucus she had watched Iraqi soldiers pull Kuwaiti babies from incubators; the testimony travelled the world, was cited by the US president, and helped sell a war — and it was theatre: "Nayirah" was the Kuwaiti ambassador's daughter, her appearance arranged by PR firm Hill & Knowlton on the Kuwaiti government's account, and the incubator story was never substantiated. Bernays' staged parade, with a war as the client. Twelve years later the machine described itself on the record: asked why the case against Iraq launched in September 2002 rather than the summer, the White House chief of staff answered, "From a marketing point of view, you don't introduce new products in August." The UK's official seven-year inquiry delivered the audit in 2016: intelligence was presented to the public "with a certainty that was not justified" (Chilcot Report). The purchase price of that certainty was paid by populations — Iraqi and Western — who were the campaign's audience, not its beneficiaries.

The stolen-election myth is the newest franchise, and its receipts are enormous. In the United States, the claim that voting machines flipped an election produced the largest media defamation payout in history — Fox's $787.5-million settlement with Dominion Voting Systems (April 2023), after discovery showed the network's own stars privately disbelieving the claims they aired — and a fundraising engine that the January 6 Committee dissected under oath: $250 million raised from small donors for an "Official Election Defense Fund" that campaign staff testified did not exist — "a marketing tactic," the campaign's own digital director agreed under oath. In Brazil, the same product line ended in court twice: the electoral tribunal barred Jair Bolsonaro from office for eight years for summoning ambassadors to push baseless voting-machine claims (June 2023), and in September 2025 the Supreme Court sentenced him to 27 years for the coup plot the myth was built to justify — the first such conviction of a former president in the country's history. Note what both cases share with every prechecked box in this article: the identity pitch ("they are stealing it from you") was also, always, a checkout flow.

And the strongman package sells on the same feeds as running shoes. In the Philippines, the drug-war presidency was marketed by a documented influence machine: Facebook's own takedown notice (March 2019) removed a 200-asset network — 3.6 million followers — organized by the man who ran the president's 2016 social-media campaign, and the journalists who mapped the operation at Rappler saw their editor share the 2021 Nobel Peace Prize. The campaign the marketing sold is now at The Hague: the International Criminal Court confirmed crimes-against-humanity charges against Rodrigo Duterte in April 2026, with trial set for November 30, 2026 (he denies the charges, saying police were instructed to kill only in self-defence). Myanmar's military ran the same play with an army of covert accounts until the UN's fact-finding mission and Amnesty concluded the platform's amplification "substantially contributed" to atrocity — the file examined in our feed investigation. Even in stable democracies the technique leaves official fingerprints: the UK statistics regulator formally rebuked the Brexit campaign's famous £350-million-a-week bus figure as "a clear misuse of official statistics" — a state auditor, correcting an ad, after the vote.

Print the counter-ledger too, because the levers are direction-agnostic. In Chile in 1988, the campaign to end the Pinochet dictatorship hired admen, built a rainbow logo and a relentlessly hopeful jingle — "joy is coming" — and used its fifteen state-granted minutes of nightly television to out-market a regime; the "No" won with about 56 percent, and the dictatorship ended at the ballot box. Identical toolkit, opposite vector. Which is the fifty-year file's real lesson: the machine has no politics — only clients. The tell, in every adjudicated case above, never changes: the message benefited its buyer, not its audience. When a message wraps your identity around someone else's agenda — your tribe, your grievance, your flag as the packaging — run the first defence question below before it runs you.

Canada 2025: the world's first news-banned election

Then Canada ran an experiment nobody designed. Because Meta responded to the Online News Act by deleting news from Facebook and Instagram for Canadians (August 2023, still in force), the April 2025 federal election was fought on platforms where journalism was structurally absent — the first national election in a G7 democracy conducted under a platform news ban. The observers measured it: the Media Ecosystem Observatory estimated roughly 11 million daily news views vanished from Canadians' feeds; in the vacuum, influencers commanded more political attention than parties or media, and hyper-partisan content expanded to fill the space; the Observatory's final judgment was that the election's integrity held — no material covert interference — while the information environment's foundations kept eroding (MEO, 2025; DFRLab). Read that against everything above: agenda-setting is the proven mechanism, and Canada handed its agenda-setting layer to whoever thrives when journalism is subtracted. The machine didn't have to persuade anyone of anything. It just changed what was on the table — which McCombs and Shaw could have predicted in 1972.

Ottawa, 2025–26: Canada's biggest feeds legally carry no news (Meta, since August 2023). The Media Ecosystem Observatory estimated roughly 11 million daily news views vanished from Canadians' feeds ahead of the 2025 election.

---

The Defence Kit That Survives Testing

Consumer defence has its own evidence standard now, and most folk advice fails it — but three defences have real results behind them, and they are learnable in an afternoon. This section is the reason the article exists; everything above was reconnaissance.

Defence one: inoculation — learn the trick before you meet it. The best-tested idea in the field is "prebunking": show people a small, weakened dose of a manipulation technique — the way a vaccine shows the immune system a pathogen — and their later detection improves. The evidence is not a lab curiosity: in seven experiments totalling about 30,000 participants, capped by a field study on five million YouTube users, 90-second explainer clips about techniques like fear-mongering and false dichotomies measurably improved viewers' ability to recognize manipulation — about a 5-percent gain in technique-recognition from a single exposure, at pennies per viewer (Roozenbeek, van der Linden et al., Science Advances, 2022; University of Cambridge/Jigsaw). Reading this article's levers section was, itself, an inoculation session. That was the design.

Defence two: the five questions. Run them on any offer, ad, or outraged post; they take thirty seconds and they target the proven levers directly. 1. Who benefits from my reaction? (the Bernays question — every staged parade has a client). 2. What is the anchor? (find the first number and ask who chose it). 3. What is the default — and who set it? (the box was checked by someone with a P&L). 4. What is missing? (the dripped fee, the "from" price, the undisclosed #ad — absence is the modern lie). 5. Would I want this if I had found it myself? (subtracts manufactured urgency, which is the only thing a countdown timer adds).

Defence three: audit your defaults on a schedule. The Iliad Flow's $2.5-billion lesson is that the machine's favourite customer is the one who never looks. Twice a year: subscriptions (every one earns its renewal or dies), notification permissions (each is a marketing channel you granted), saved payment methods (one-click is a lever aimed at you), and the "angry" reactions your feed has learned to farm — a mechanism this publication documented from the platforms' own files in the feed investigation. For purchases over a set line, adopt the 30-day list: if it still matters in thirty days, buy it deliberately. Scarcity theatre cannot survive a month.

The audit that beats the machine (staged illustration): scheduled default checks — subscriptions, permissions, saved cards. The Iliad Flow's $2.5-billion lesson is that the maze only holds people who never look.

And use the machinery built for you. In Canada, drip pricing and fake discounts are now Competition Act violations — the Competition Bureau takes complaints, and the Cineplex judgment proves they can end in nine-figure penalties; misleading or undisclosed influencer ads can be reported to Ad Standards under the Canadian Code; broadcast complaints go to the CRTC; unwanted commercial email violates CASL. For the political layer: when a message enrages you on schedule, ask the McCombs question — who put this on my agenda? — and check whether journalism or an engagement algorithm chose your morning's outrage. In a country whose feeds legally carry no news, that question is not rhetorical.

---

Zeus Shows Its Hand

We sell electric bikes. That makes us participants in the industry this article just spent ten thousand words dissecting, and by our own method — judge behaviour, not claims — a disclosure section is the only honest way to close. So here is our hand, stated so you can audit it.

The levers we use, in the open: we publish long, sourced buying guides because informative content earns trust and trust earns customers — that is marketing, and it is this article. We show real prices with financing math beside them, we cite our sources inline, and we tell you when a bike is wrong for you because honesty measurably reduces returns and builds the only asset we care about. The levers we have banned, in writing, in our own editorial rules: no manufactured scarcity or fake countdowns (real inventory limits only, verifiable against our stock), no drip pricing, no status-escalation flattery ("insider," "elite," "what they don't want you to know"), no competitor-bashing in social posts, and no invented statistics, ever — every number sourced or absent. This page contains no product links because you cannot honestly anatomize the machine while operating it on the reader. If you ever catch a Zeus page violating this list, the complaint routes in the section above work on us too — and we would rather hear it directly: 1-866-938-7580, a human answers.

---

FAQ: What Canadians Ask About Marketing — Answered From the Record

Is marketing manipulation or value creation?

Both exist, and the honest test is observable: value-creating marketing survives full information (clear price, real reviews, easy cancellation), while manipulation depends on hiding something - a dripped fee, a prechecked box, a fake discount. Courts now draw the same line: Amazon paid $2.5 billion in 2025 over a cancellation maze, and Cineplex paid $38.9 million over a hidden $1.50 fee. Information is marketing; engineered misunderstanding is the trick.

Does subliminal advertising actually work?

The famous version is a confessed hoax. James Vicary, who claimed in 1957 that flashing 'Eat Popcorn' frames boosted cinema sales 57.5%, admitted in a 1962 Advertising Age interview that the experiment was a gimmick invented to save his failing business, and rigorous replications found no meaningful effect. What does work is entirely visible: anchoring, defaults, scarcity cues, and engineered feeds - tricks that operate in plain sight.

Is subliminal advertising illegal in Canada?

In broadcasting, yes - Canada's Television Broadcasting Regulations, 1987 prohibit broadcasting anything that uses a subliminal device. The irony is that the ban outlived the evidence: the founding subliminal experiment was a confessed fabrication, and replications failed. The techniques that measurably move consumers - defaults, drip pricing, urgency cues - have never been subliminal. They work in plain sight, which is why modern law targets them instead.

How does advertising manipulate consumers?

The documented levers are few and old: anchors (a fake 'was' price resets your sense of value), defaults (prechecked boxes decide for you - organ-donor consent roughly doubles under opt-out), scarcity and urgency cues (countdown timers), social proof ('37 people are viewing this'), and variable rewards (loot boxes use the same reinforcement schedule as slot machines). Each is proven by measured behaviour - purchases, sign-ups, refund floods - not by marketer claims.

Why do we buy things we don't need?

Mostly because the choice environment was built by professionals and your defaults were set for you: anchored prices make anything look like a bargain, engineered urgency compresses deliberation, one-click flows remove the pause where second thoughts live, and identity marketing ties products to who you want to be. The century-old proof is Bernays' 1929 'Torches of Freedom' stunt - cigarettes sold to women as liberation, and women's share of cigarette sales more than doubled by decade's end.

What are dark patterns?

Interface designs engineered to trick you - prechecked boxes, hidden fees, cancellation mazes, disguised ads, forced urgency. The term is now regulatory language: the FTC fined Epic Games $520 million in 2022 partly for dark patterns, and in September 2025 Amazon paid $2.5 billion over its 'Iliad Flow' - a four-page, six-click, fifteen-option Prime cancellation maze. In Canada, the Competition Act's 2022-2024 amendments made the most common dark pattern - drip pricing - explicitly illegal.

What is drip pricing - and is it illegal in Canada?

Drip pricing is advertising a price you cannot actually pay, then 'dripping' mandatory fees into checkout. It is now explicitly against Canada's Competition Act (2022 amendments, strengthened June 2024) unless the added charge is a government tax. The landmark case is Cineplex: a mandatory $1.50 online booking fee cost the company a $38.9-million penalty - roughly every dollar the fee collected - upheld by the Federal Court of Appeal in January 2026.

How are boys and girls targeted differently by marketing?

Deliberately, and it changed within living memory: 1970s catalogues consciously de-gendered toys, LEGO's famous 1981 ad showed a girl in jeans with the caption 'What it is is beautiful' - then the industry re-segmented in the 1980s-90s because two gendered aisles sell more than one. Content analyses find boys' marketing sells skill and heroism while girls' sells appearance and caretaking. The bill arrives in adulthood as the pink tax: New York's audit of about 800 products found women's versions cost more 42% of the time (men's: 18%).

Did Cambridge Analytica really swing elections?

The forensic verdict says the legend outran the evidence. The UK Information Commissioner seized the company's servers and concluded in 2020 that its models were 'exaggerated' hype, and that it did not work on the Brexit referendum beyond initial inquiries. Randomized experiments agree: across 59 trials, political ads moved candidate preference only slightly regardless of message or targeting. The data harvesting was real and wrong; the mind-control story was marketing - by the manipulation industry, about itself.

What propaganda techniques were used in the world wars?

WWI built the modern template: the US Committee on Public Information (the Creel Committee) deployed roughly 75,000 volunteer 'Four Minute Men' who delivered over 7.5 million short speeches in cinemas and halls, alongside posters, films and press placement - atrocity framing, fear appeals, social pressure and testimonial. The methods then moved into commerce: Edward Bernays, who worked with the committee, applied the same playbook to cigarettes, soap and breakfast within a decade.

Do political attack ads work?

Sometimes devastatingly, sometimes they detonate on the sender - and Canada owns the clearest example of each. The 1993 Progressive Conservative ad that appeared to mock Jean Chretien's facial paralysis backfired so badly the party fell to two seats. The 2011 'Just Visiting' campaign against Michael Ignatieff defined him and worked. Randomized experiments find the average persuasion effect of any single political ad is small; the damage comes from sustained definition of a candidate, not one spot.

Has marketing ever been used against an entire population?

Yes, and the receipts are judicial. A UN tribunal convicted three Rwandan media executives in 2003 of inciting genocide - the first such convictions since Nuremberg - and a Quarterly Journal of Economics study attributed roughly 51,000 perpetrators (about 10% of participation) to villages inside the radio station's signal. Wars have been sold with fabricated testimony arranged by a PR firm (the 1990 'incubator' hoax), and stolen-election myths produced history's largest defamation payout ($787.5 million) plus a $250-million fund that did not exist. The tell never changes: the message benefits its buyer, not its audience.

How do I stop falling for marketing tricks?

Use the defences that survived testing: inoculation ('prebunking') measurably improves manipulation-detection - in a 5-million-user YouTube field study, one 90-second explainer clip improved recognition of manipulation techniques by about 5% - so learn the levers before you meet them. Then run the five questions: Who benefits from my reaction? What is the anchor? What is the default? What fee or fact is missing? Would I buy this if I had found it myself? And audit your subscriptions - the Iliad Flow only works on people who never look.

---

Sources & Further Reading

Grouped by section; primary institutions and peer-reviewed work first. All links verified live July 26, 2026.

The billed week & dark patterns: US FTC — the historic $2.5B Amazon settlement (Sept 2025) · National Law Review — the settlement's dark-pattern focus · Competition Bureau — Cineplex drip-pricing win · Canadian Advertising Law — FCA upholds the $38.9M Cineplex decision (Jan 2026) · Competition Bureau — guide to the June 2024 Competition Act amendments · European Commission — CPC Network notification to Temu (Nov 2024) · Ad Standards — Influencer Marketing Disclosure Guidelines

The invention (1917–1929): Library of Congress — the Four Minute Men · Four Minute Men — program records · History Today — Bernays, the original influencer · University of Ottawa digital archive — the Torches of Freedom campaign · UCSF Truth Tobacco Industry Documents — the 1969 "Doubt is our product" document · The Marlboro Man — from "Mild as May" to the cowboy (1954) · Cowboy State Daily — the Marlboro Men history

The hoax & the levers: James Vicary — the 1962 Advertising Age admission · Johnson & Goldstein (2003) — "Do Defaults Save Lives?", Science 302 · Anchoring — Tversky & Kahneman's 1974 wheel demonstration · APS — the hotel-towel field experiments (Goldstein, Cialdini & Griskevicius, 2008) · Schimmack, Heene & Kesavan (2017) — the priming "train wreck" analysis, with Kahneman's reply · Kahneman's 2012 open letter urging replication · Maier et al. (2022) — "No evidence for nudging after adjusting for publication bias", PNAS · Xiao (2023) — Belgium's loot-box ban audit, Collabra

The Segment Files: Ali et al. (2019) — "Discrimination through optimization" (Facebook ad-delivery skew) · CNBC — DOJ–Meta Fair Housing settlement (2022) · NYC DCA — "From Cradle to Cane" gender-pricing study (2015) · The Conversation — research review on gendered toy marketing · Flashbak — LEGO's 1960s–80s gender-neutral marketing archive · AGCO — Ontario bans athletes from igaming ads (eff. Feb 2024) · EFF — the WinRed prechecked-box scheme · US DOJ — the $650M McKinsey–Purdue resolution and the "turbocharge" engagements (Dec 2024)

The Rulebook Atlas: University of Illinois — Dhar & Baylis on Quebec's child-ad ban (JMR) · Health Policy Watch — Chile's food-marketing law and its results · STAT — Chile's honest caveat: consumption fell, obesity didn't · Australian Government — plain-packaging Post-Implementation Review (2016) · Croud — China's 2015 Advertising Law: the banned superlatives · Law.asia — China's banned-language enforcement practice · EPHA — Russia's 2012 alcohol-advertising ban · European Journal of Public Health — modelling child-ad bans and obesity

The political wing: Britannica — the Daisy ad · The Marshall Project — Willie Horton revisited, with Atwater's Life-magazine apology · Frank Luntz — the memos and the 2019 recantation · The 1993 Chrétien attack ad · "He Didn't Come Back For You" — the 2011 Ignatieff campaign · MarketingSherpa — the Obama 2012 email-testing program · NYRB — the 5.9-million-variant Facebook operation · NBC — the FEC's 6–0 response to prechecked donation boxes · Computer Weekly — the ICO closes the Cambridge Analytica investigation · Coppock, Hill & Vavreck (2020) — 59 randomized ad experiments, Science Advances · Aggarwal et al. (2023) — the 2-million-person campaign field experiment · First Draft — 100,000 WhatsApp messages from Brazil's 2018 election · HKS Misinformation Review — WhatsApp images in India's 2019 election · Media Ecosystem Observatory — Canada's 2025 information ecosystem · DFRLab — how social media shaped the 2025 Canadian election · NY Attorney General — 18 million fake FCC comments (2021) · TechCrunch — 80% of net-neutrality comments were fake · UN ICTR — three media leaders convicted of genocide (Dec 2003) · Yanagizawa-Drott (2014) — "Propaganda and Conflict", QJE (open access) · The Nayirah testimony — the Hill & Knowlton incubator fabrication (1990) · Democracy Now archive — Andrew Card's "marketing point of view" (Sept 2002) · UK Parliament — the Chilcot Report of the Iraq Inquiry (2016) · NBC — the $787.5M Dominion–Fox settlement (2023) · NPR — the January 6 Committee on the nonexistent Election Defense Fund · PBS — Brazil's electoral court bars Bolsonaro for baseless voting-machine claims (2023) · NPR — Bolsonaro sentenced to 27 years for the coup plot (Sept 2025) · Meta newsroom — removing the Gabunada network in the Philippines (2019) · International Criminal Court — the Duterte case · UK Statistics Authority — the £350M figure: "a clear misuse of official statistics"

The defence kit: University of Cambridge — the 5-million-user prebunking field study · Traberg, Roozenbeek & van der Linden (2022) — psychological inoculation: current evidence · Competition Bureau Canada — complaints · Ad Standards Canada

---

The Bottom Line

A hundred years of receipts reduce to three sentences. The persuasion machine is real, old, and architectural — it wins through anchors, defaults, mazes, segment files and agenda control, not through hidden frames or mind-reading, and its most mythologized powers (subliminal messages, psychographic mind control) are, on forensic inspection, its own best advertising. It works identically on soap, on children's games, on your grandmother's donations and on your ballot — and when countries have dared to switch pieces of it off, from Quebec to Canberra, behaviour measurably changed, which is the proof of both its power and its governability. And the defence is neither cynicism nor retreat: it is literacy — the levers learned in advance, the five questions asked on schedule, the defaults audited — because a public that can read the machine is the only regulator that scales.

The civic ask, in the spirit of everything above: send this to one person who shops, one person who scrolls, and one person who votes — they are the same person, and the machine already knows it. The rest of our ledgers continue the argument: the hundred documented lies, the feed that farms your anger, the country that regulates the feed, and the bill Canadians pay for concentrated markets.

📸 All photography by Playcut.ai — personalized AI actor technology