Pick up a box of Froot Loops (or any children's cereal). What you see: Toucan Sam, a character designed to make children feel playful excitement and brand familiarity. Bright primary colours (red, yellow, blue) — research shows these stimulate appetite and excitement in children. The word "fun" appears multiple times. A claim about vitamins and minerals implies the cereal is healthy. A bright starburst saying "New!" implies novelty.

None of this happened by accident. A team of designers, marketers, and child psychologists made deliberate choices about every element. This is not just a package — it is an advertisement that your child encounters before they are even old enough to know what advertising is.

A popular gaming creator says: "This video is brought to you by [VPN company]. I've actually been using it for the last six months and I genuinely recommend it — use code [X] for 30% off." This is advertising. The creator has been paid, sometimes thousands of dollars per mention. By law they must disclose this — but studies show that many disclosures are buried in descriptions, shown very briefly, or framed to feel like casual personal recommendations. The warmth and familiarity of the creator's voice is the product's advertising strategy.

This classic advertising formula sounds like a fact (it has a number) but conceals almost all the relevant information. Which 5 dentists? Compared to what alternative? Who paid for the survey? How was "recommend" defined? Was it "recommend above all others" or "would not object to"? The claim has the structure of evidence but the content of almost none.

Technically a fact — if you also eat eggs, toast, fruit, and drink orange juice, then adding this cereal makes the breakfast balanced. But the image always shows just the cereal prominently. The fact is technically true; the impression created is that the cereal itself is a balanced choice. This is a masterclass in using true facts to create false impressions.

Compare: "The government announced new housing policy." (Fact — you can check whether this announcement happened.) Versus: "The government's disastrous housing policy will hurt millions." (Opinion embedded in a factual-sounding sentence — "disastrous" and "will hurt" are judgements, not checkable facts.) Children learn to spot the evaluative language that turns a factual claim into an opinion.

Claim: "Scientifically proven to improve your child's concentration."

Who said it? The company selling the cereal. They have an obvious financial interest in you believing this claim.

Can we check it? The claim cites "science" but doesn't point to any specific study. A real scientific claim would cite a peer-reviewed paper that independent researchers could read and evaluate. Without that, "scientifically proven" is a phrase, not evidence.

Does it make sense? A breakfast cereal improving concentration is plausible if it provides energy and reduces hunger — but "improving concentration" is vague. Compared to what? By how much? In which children? Under what conditions? The claim sounds more specific than it is.

Claim: "A new study shows that [popular thing] causes [scary health outcome]."

Who said it? The post doesn't link to the study. The account sharing it has a history of sharing alarming health content. Red flag.

Can we check it? Search for the actual study. Does it exist? Was it published in a reputable peer-reviewed journal? Was it conducted on humans or animals? How large was the sample? Was it widely reported by science journalists? Often the "study" is either misrepresented, not peer-reviewed, or conducted with a sample of 12 people.

Does it make sense? If this were genuinely a major health finding, it would be reported by science reporters at major news organisations — not just by viral social media posts. The absence of mainstream scientific coverage is itself a signal.

"Clinically proven to help you lose weight fast — no diet required, no exercise needed!" This claim stacks multiple red flags: "clinically proven" (by whom? published where?), "fast" (what counts as fast?), "no diet or exercise" (contradicts everything we know about how bodies work). The UK's Advertising Standards Authority regularly sanctions supplement advertisers for claims exactly like this — and yet they persist because the emotional appeal of effortless weight loss is powerful enough that people buy first and scrutinise later.

"Boost your IQ by 30 points in just 21 days!" Brain training apps have made versions of this claim for years. In 2016, Lumosity was fined $2 million by the US FTC for deceptive advertising after claiming its app could prevent cognitive decline and improve performance in everyday life — claims the company had no scientific evidence to support. The claim felt plausible and the emotional appeal was powerful. That combination is exactly when the "too good to be true" alarm should ring.

"LIMITED OFFER — only available for the next 4:32!" The timer creates urgency — a feeling that you are about to lose something. That feeling (anxiety + FOMO) is specifically engineered to bypass deliberate thinking and make you act before you've evaluated whether you actually want the thing. The pause button: notice the urgency feeling. Ask: will this offer actually expire, or will it reset? Do I want this thing independent of the timer? Am I making this choice or is the timer making it for me?

"[Group of people] are DESTROYING [beloved thing]!" This headline is engineered to generate outrage — a feeling powerful enough that millions of people share content they've never read past the title. The outrage arrives first; any evaluation of whether the claim is true comes second, if at all. Pause: I notice I feel angry. That anger was the point of this headline. Now let me read the actual article and see if the story supports the headline.

The manufacturer's motive: sell as much cereal as possible, as profitably as possible. That motive explains every single design choice. The cartoon character: builds brand loyalty in children who then influence parental purchasing. The health claim: counters parental scepticism. The bright colours: stimulates desire in children at the point of sale. Every element serves the financial motive — understanding that doesn't mean the cereal is bad, but it means the packaging is not neutral information about the product.

A creator loves the game (genuine motive: share enthusiasm). The video also generates advertising revenue (financial motive: impressions = income). The creator's audience expects positive content about their favourite games (social motive: maintain relationship with audience). The creator may have received early access from the developer (relationship motive: maintain access). None of these motives are necessarily sinister — but they all influence what the video says and doesn't say. A genuinely critical review might damage all four relationships simultaneously.

Have you ever ordered something from a menu that looked dramatically better in the photo than it did on the plate? Food photography is a highly specialised skill that uses techniques including: glue instead of milk (to keep cereal from going soggy), motor oil instead of syrup (for a glossy look that lasts under hot lights), paint and blowtorches to achieve perfect grill marks on meat, and ice cream made from coloured shortening that won't melt under studio lights. Every food photo is a constructed image. This isn't illegal — it's just important to know when looking at the picture.

A photograph of a protest can be cropped to show a crowd of thousands, implying massive public support — or cropped to show a small group, implying fringe interest. Both crops can come from the same photograph, taken at the same moment. Neither is "false" in the technical sense — but they tell entirely different stories. The same technique applies to polling data, statistics, and quotations.

In 2023, the US Federal Trade Commission sent warning letters to major brands including Sony, Walmart, and Universal Music Group, as well as numerous individual influencers, for failing to adequately disclose paid partnerships. Many influencers had used phrases like "thanks to [brand] for sending me this" or included #ad in a list of other tags rather than prominently at the start of a post. The FTC's guidance is clear: disclosures must be hard to miss. The frequency of enforcement actions reveals how routinely this standard is not met.

In 2014, Microsoft paid the gaming network Machinima to have YouTubers promote the Xbox One without disclosing that the content was paid advertising. Machinima paid creators a bonus rate per view on Xbox promotional content — but creators were instructed not to mention the sponsorship. The FTC investigated and Machinima was required to implement a disclosure programme. This is a documented case of an entire content network being used as undisclosed advertising infrastructure.

Apply the two-score method to these real headlines. "You Won't BELIEVE What This Dog Did!" — Emotion: 5/5. Information: 0/5. The headline tells you nothing except that you should be surprised. "Dog rescues owner from house fire by barking until neighbours called emergency services." — Emotion: 3/5 (the story is genuinely touching). Information: 5/5. The second headline is less compelling to click — but if you had to share one with a friend, which would you rather send them?

YouTube thumbnails have become increasingly extreme because creators compete for attention in a scroll. Studies of YouTube thumbnails show a clear evolution over time: more extreme expressions (wide eyes, open mouths), more text in larger fonts, more saturated colours. Many creators have acknowledged the "thumbnail game" — the pressure to make thumbnails that promise more drama than the content delivers. Some creators have publicly spoken about how exhausting and dishonest the format feels, while continuing to use it because the alternative is fewer views.

"Recommended by leading dermatologists." Questions: Which dermatologists? How many? Were they paid to recommend it? What does "recommended" mean — preferred over everything else, or merely not objecting to its use? This phrasing is deliberately designed to invoke the authority of medical expertise without meeting the evidential standard that expertise would actually require.

Political advertising routinely employs false dichotomies: "Either you're tough on [issue] or you're for [terrible thing]." This framing eliminates every nuanced position between the two extremes. Once you recognise the structure, you see it in almost every political advertisement — and it becomes much less persuasive.

In 2018, a California court ruled that coffee sellers must post cancer warnings after a case citing a 2016 study. The headlines were alarming: "Coffee causes cancer, court rules." The primary source: a study of occupational acrylamide exposure (a chemical in many roasted foods, not just coffee) in populations exposed to far higher doses than typical coffee consumption. The scientific consensus, per the World Health Organisation, is that coffee is not classifiably carcinogenic and may actually be protective against some cancers. The court ruling was a technicality of California's Proposition 65 law, not a statement of scientific consensus. The headline and the primary evidence told entirely different stories.

When a dramatic image circulates on social media — a flooded city, a dramatic protest, a shocking moment — reverse image search takes you to the primary source: when and where the image was originally published. This frequently reveals that the image is from a different country, a different event, or a different year than the post claims. Tools: Google Images reverse search, TinEye. Time required: approximately 30 seconds.

Many games use a "Battle Pass" or seasonal content model: exclusive items, characters, or storylines available only during a limited window. The framing: "Season ends in 23 days — don't miss out!" The emotional mechanism: items tied to social status in the game (other players will see what you have), time-limited availability creating scarcity anxiety, and the sunk cost of the game investment making FOMO more acute. The design is built around FOMO as a purchase driver, not around the value of the items themselves.

MIT research published in Science (2018) analysed 126,000 news stories on Twitter over 11 years and found that false stories spread to more people, more quickly, and more deeply than true stories in every category of news. The primary driver: false stories tended to be more novel and emotionally arousing — especially generating surprise and disgust — than true stories. People shared false stories not because they were gullible but because the emotional response to novelty and moral outrage is a powerful, fast-acting drive to tell others.

A free mobile game allows players to progress easily for the first few levels, then introduces a difficulty spike — a point where progress becomes frustratingly slow. At exactly this moment, a "starter pack" appears: "Get 5x resources + exclusive character for just $0.99!" The offer is timed to coincide with peak frustration. The $0.99 price point is deliberately low — below the threshold most people apply deliberate financial reasoning to. Many games layer multiple such purchases, each individually trivial, to generate significant total spending.

In 2022, the US FTC published a report specifically on dark patterns, warning companies that deliberately confusing or manipulative UX design may violate consumer protection laws. Amazon was subsequently sued over its Prime cancellation flow — and in September 2025 settled with the FTC for $2.5 billion, one of the largest consumer protection settlements in history. Internal Amazon documents revealed during the case showed employees openly describing their own enrolment tactics as "a bit of a shady world." This is the clearest public evidence that dark patterns are not accidents but engineered features.

YouTube's CEO disclosed in 2018 that the recommendation algorithm drives approximately 70% of the total time users spend on the platform. This means that for every 10 minutes a typical user spends on YouTube, 7 of those minutes are watching something the algorithm chose for them, not something they searched for. The platform is not a search engine for content you want — it is a recommendation engine serving content the algorithm believes will keep you watching.

The Wall Street Journal conducted an experiment in 2021: they created bot accounts on TikTok and had them behave in specific ways (pausing on certain content, watching certain topics in full). Within hours, the accounts were receiving highly targeted content based purely on viewing behaviour. Accounts that engaged with content about depression or body image received increasing amounts of related content. The algorithm had no way to evaluate whether this content was good for the user — it only knew it generated engagement.

In 2022, the UK's Competition and Markets Authority investigated Amazon and other platforms over fake reviews. Research by Which? found that a single search on Amazon for popular products returned results where a significant proportion of top-ranked products had review profiles showing classic signals of manipulation: highly clustered review dates, very high five-star percentages inconsistent with the category average, and generic positive language. Amazon has removed hundreds of millions of fake reviews — and hundreds of millions more likely remain.

A product with 10,000 reviews showing 95% five-stars and almost no 2-3 star reviews is more suspicious than one with 80% five-stars and a natural distribution. Real products have a bell curve of opinions. A suspiciously perfect distribution suggests either selection bias (only satisfied customers were prompted to review) or fake reviews. Also look at the text of the five-star reviews: do they read like genuine experience or like promotional copy?

In 2021, documents provided to the US Congress and Wall Street Journal by whistleblower Frances Haugen included internal Facebook research showing that the company was aware that Instagram was harmful to a significant proportion of teenage girls' mental health — and had internal evidence that its own algorithms amplified divisive and extreme content because it drove engagement. The documents showed that Facebook considered and largely shelved changes that would have reduced harmful content because they projected it would reduce engagement metrics. The business model created a direct conflict between user wellbeing and revenue.

Facebook earns approximately $50 per US user per year in advertising revenue. The average US Facebook user spends approximately 30 minutes per day on the platform, or approximately 180 hours per year. This means the platform earns roughly $0.28 per hour of your attention — which it sells to advertisers. When you understand this, the design of every feature — the notification, the autoplay, the "people you may know" — becomes clearly readable as an engineering decision about how to extract more of those hours.

The 1998 Wakefield study falsely linking vaccines to autism was retracted in 2010, and Wakefield was struck off the medical register for ethical violations. Despite this, vaccine hesitancy persists in many communities. This is a case study in confirmation bias at scale: people who had already formed the belief that vaccines were harmful continued to find and share the supporting evidence (the retracted study) while applying intense scrutiny to and dismissing the overwhelming contrary evidence (decades of epidemiological research). The belief had become identity-linked, making motivated reasoning particularly powerful.

Psychologist Dan Kahan conducted experiments in which participants were given data about a controversial policy question (gun control, climate change, nuclear waste) and asked to evaluate the data's conclusions. Participants consistently interpreted identical data as supporting their prior political position — the same dataset "showed" opposite things to liberals and conservatives. Crucially: the effect was stronger among participants with higher numeracy. More analytical people were better at finding ways to make the data say what they already believed.

Whataboutism is one of the most frequently deployed rhetorical techniques in political media. Its structure: a criticism is made of Group A; a supporter of Group A responds by pointing to a failing of Group B. The technique is effective because it appears to engage with the criticism while actually doing nothing but change the subject. The original criticism is left unaddressed. Identifying whataboutism requires asking: does the counterexample actually bear on the validity of the original criticism? In almost all cases, the answer is no.

A common journalistic technique of presenting "both sides" on contested questions can slide into false equivalence when the two sides are not equally evidence-based. Covering the scientific consensus on a topic alongside a fringe contrarian view implies an equivalence of evidential weight that doesn't exist. This can also become a straw man: if the contrarian's weakest arguments are selected for the "against" side, the debate appears easier to resolve than it actually is.

In early 2020, a false claim linking 5G mobile networks to COVID-19 spread rapidly across social media platforms, resulting in the arson of mobile phone towers in the UK and violence against telecommunications workers. The claim had no evidential basis — viruses are biological; 5G is electromagnetic radiation; the two cannot interact. Yet the claim spread for several reasons: it was novel, it was emotionally arousing (fear + conspiracy), it was shared by accounts with large followings, and it connected to pre-existing anxiety about both the new technology and the pandemic. The correction — simply factual — spread slowly and to a smaller audience.

The Cambridge Social Decision-Making Lab created a free online game called "Bad News" that teaches players the techniques used to create misinformation: impersonation, emotional manipulation, polarisation, trolling, conspiracism, and discrediting. A study of 15,000 players found that playing the game significantly improved players' ability to identify and resist manipulative content — and the effect persisted over time. This is experimental evidence for the core premise of Early Decoder: knowing the techniques is protective against them.

The Atlantic published a native advertising piece from the Church of Scientology that was formatted identically to editorial content, with the "Sponsor Content" label almost invisible. Comments critical of Scientology were deleted from the article. When readers noticed, the backlash was severe enough that The Atlantic took the piece down and apologised. This case became a landmark in discussions about native advertising ethics — and a clear example of what the format looks like at its most problematic.

Look for: a small "Sponsored," "Paid post," or "Partner content" label, usually in a lighter colour and smaller font than the headline. Check whether the article's topic suspiciously matches a company's product or service. Look at whether critical perspectives on the sponsor are included (they almost never are in native advertising). Check if clicking the author's name leads to a company page rather than a journalist's profile. Compare the URL structure to regular articles on the same site.

Cambridge Analytica, a political data consultancy, harvested the Facebook data of approximately 87 million users without their consent through a personality quiz app. The data was used to build psychographic profiles of American voters, which were then used to micro-target political advertising in the 2016 US presidential election — serving different emotional appeals to different psychological profiles. The UK's Information Commissioner's Office fined Facebook £500,000 (the maximum under then-current law) for its role in the data breach. This case is the clearest documented example of psychological profiling being used for large-scale political persuasion.

The phenomenon of searching for a product and then seeing ads for it everywhere for weeks is called retargeting. It is not coincidence or coincidence — it is a deliberate advertising mechanism. Your browsing activity on one website is tracked by third-party advertising networks, which then purchase ad slots on other websites to show you ads for the product you searched for. The experience of being "followed" across the internet by an ad is a visible manifestation of the invisible data-collection infrastructure that underlies the entire commercial web.

In the US and UK, violent crime rates fell substantially from the early 1990s through the 2010s. During the same period, crime coverage in news media increased. Surveys consistently show that majorities of the public believe crime is rising in their country even when it is statistically falling. This is a textbook case of negativity bias and normalisation bias in action: crime that happens is news; crime that doesn't happen isn't. The structural bias in coverage creates a systematic gap between the statistical reality and public perception.

Rupert Murdoch's News Corp owns media assets across the US, UK, and Australia including Fox News, The Wall Street Journal, The Sun, and The Australian. Research into the political alignment of these outlets consistently finds them aligned with Murdoch's documented political preferences. No evidence of direct editorial instruction is required to explain this alignment: editors are hired by owners, editorial culture is established over time, and journalists at ideologically consistent outlets self-select. The alignment is real; it's produced by structural forces rather than individual corruption.

Early in the Russian invasion of Ukraine, a deepfake video circulated appearing to show Ukrainian President Volodymyr Zelensky announcing that Ukraine was surrendering and urging soldiers to lay down their arms. The video was quickly identified as a deepfake — the face looked slightly unnatural, the neck proportions were wrong, and the voice was inconsistent with Zelensky's known speech patterns. Zelensky rapidly posted an authentic video in response. This is a real documented case of deepfake video being deployed as active warfare disinformation.

Researchers at the Stanford Internet Observatory and NewsGuard have identified hundreds of websites that publish AI-generated news content at scale — with no human journalists, operating across multiple countries and topics, generating advertising revenue from the volume of content. The articles are often superficially plausible but contain factual errors, missing attribution, and inconsistencies that careful reading reveals. They are designed to be good enough to share, not good enough to scrutinise.