While Brussels has begun fining fake content and Brasilia is still debating the fine print of its legislation, YouTube’s algorithms have already found the ideal victim: people over seventy
An 82-year-old Brazilian woman once came across a YouTube video in which a man in a white coat calmly explained how diet could help preserve eyesight in old age. She believed him immediately and without reservation.
The platform had clearly labeled the video as created with artificial intelligence, with a warning displayed directly above the player. The woman either failed to notice the label or paid no attention to it: she was convinced she was watching a real doctor.
The doctor, however, did not exist. His face, voice, and manner of speaking had been synthesized by an algorithm, while the channel that posted the video was making money not by treating patients but by converting their anxiety into advertising impressions.
Behind this individual video lies neither an accident nor an eccentric enthusiast experimenting with neural networks, but a standardized, reproducible commercial model that can be deployed just as easily in any language and for virtually any disease.
70 Million Views: Fake Medicine Has Become an Industry
This case is not an anomaly. It is a symptom of an entire industry.
An investigation by CTRL+Z, a Brazilian nonprofit specializing in data journalism, identified 29 Portuguese-language YouTube channels using AI-generated “doctors” that had accumulated at least 70 million views in total.
A parallel review found at least 50 instructional videos in nine languages, Portuguese, English, Hindi, Spanish, French, Italian, Polish, German, and Turkish, explaining in detail how to build such a channel from scratch.
The scheme has long since crossed the boundaries of any single country or language. It has become an export product, a disinformation franchise that can be localized for any retirement-age audience anywhere in the world.
Charlatans Have Always Existed. Now They Do Not Even Need a Real Person
The scheme itself is nothing new.
The patent-medicine industry flourished in the English-language press at the turn of the nineteenth and twentieth centuries, when newspaper advertisements sold supposed elixirs for cancer and tuberculosis with virtually no scrutiny. Radio in the 1930s produced its own “radio doctors,” broadcasting on the strength of dubious medical credentials until U.S. federal regulators began revoking their licenses on a large scale.
Every new communications technology, newspapers, radio, television, and now generative artificial intelligence, goes through the same phase: a period in which the speed at which content spreads outpaces the development of social and legal safeguards against abuse.
The best-known example from the last century was American radio doctor John Romulus Brinkley, who in the 1930s used one of the era’s most powerful border radio stations to promote rejuvenation surgeries of his own invention despite lacking a medical education. He managed to build a fortune and reputation comparable to those of a major pharmaceutical company before regulators finally stripped him of his license.
The difference is that yesterday’s charlatans needed at least one real person with a reasonably plausible backstory. Today’s version of the scam can be produced without a single real actor: the “doctor’s” face, voice, and biography can be assembled by an algorithm in minutes and reproduced simultaneously in nine languages.
First They Frighten You With Death. Then They Sell You Ginger and Supplements
The mechanics of these videos are almost always the same.
A figure in a white coat, sometimes entirely synthetic, sometimes a digital replica of the face of a real and unsuspecting physician, announces that an ordinary food, medication, or everyday habit is secretly killing the viewer.
Then comes the promise of a miraculous and supposedly forgotten remedy: ginger, sweet potatoes, or some particular combination of herbs.
One creator of training courses for people building such channels publicly claimed that this type of business could generate as much as 30,000 Brazilian reais, approximately $5,800, per month.
Monetization comes from three revenue streams: YouTube advertising, sales of electronic books, and affiliate links to “natural” supplements.
The same course creator showed students how to intensify the alarmist tone and pressure viewers into taking immediate action, even using language suggesting that they might not wake up the next morning.
The attention economy works almost flawlessly here because fear of death is a far more effective driver of clicks than any rational recommendation.
Who Profits From a Patient’s Fear
Following the money to the end of the chain reveals a system far more complex than a lone scammer with a laptop.
YouTube receives its share of advertising revenue regardless of whether the doctor on screen is real or synthetic. Until a complaint is filed, its ranking algorithm promotes both types of videos with equal enthusiasm because it measures engagement rather than credibility.
Voice-cloning and face-generation services on which these channels are built generally bear no responsibility for how their products are used. Their business model is based on subscriptions, not on moderating their customers’ content.
At the end of the chain are dietary-supplement vendors and electronic-book authors, whose profits depend directly on how frightening the headline produced by the previous link in the chain turns out to be.
Each participant in this chain can formally claim innocence in isolation, while having every financial incentive to keep the system functioning as an integrated whole.
The Fear Playbook Is Sold Separately
The story of one creator of these training courses is revealing in its own right.
After journalists contacted him, public access to one of his demonstration videos was restricted. The creator said the channels shown were merely examples and did not belong to him personally. According to him, he teaches people how to create informational content using artificial intelligence, not material intended to replace medical consultations, and he never meant to encourage violations of platform rules.
Formally, that may well be true. A course about making money on YouTube is not legally identical to a specific fraudulent video.
But a manual explaining how to heighten a viewer’s anxiety by warning that they may not wake up in the morning works regardless of who eventually uses it. The instructor profits from selling the course, while moral and legal responsibility for the final outcome is dispersed among dozens of anonymous students, each of whom can formally claim to have acted independently.
A $220 Billion Market: Why “AI Doctors” Have Appeared Right Now
The industry into which the revenue ultimately flows is anything but marginal.
According to industry research, the global dietary-supplement market is expected to reach roughly $220 billion in 2026 and continue growing by 7 to 9 percent annually until at least the middle of the next decade, driven in large part by demand for healthy-aging products.
Synthetic doctors do not create this demand. They parasitize it. They exploit an already multibillion-dollar consumer appetite for alternatives to conventional medicine and add just one new layer: a fabricated authority figure in a white coat, produced by an algorithm at almost no cost.
Why Older People Have Become the Primary Target
The choice of audience is not accidental.
The share of people over sixty in the global population has been rising steadily for decades, and the advertising industry has long recognized this as a structural shift in demand rather than a temporary trend.
According to United Nations estimates, by 2050 one in six people worldwide will be over the age of sixty-five, nearly twice the proportion seen today.
For the creators of synthetic medical channels, this is not abstract demography. It is a direct signal about where to invest. An audience now considered a niche will, within a generation, become one of the largest groups of video-content consumers in the world.
Older people spend more time on video platforms than is commonly assumed, are less likely to cross-check what they see, and are more likely to have disposable income that can be redirected toward purchasing an advertised supplement.
Creators of this type of content openly describe older audiences as a profitable niche in their training materials, not because these people are inherently naive, but because they have time, money, and, most important for a fraudulent sales funnel, an inherited trust in the authority of a doctor in a white coat, built over decades of experience with traditional, in-person medicine.
An 85-Year-Old Patient Replaced His Medication With Sweet Potatoes
The consequences are not merely theoretical.
An analysis of roughly 27,000 comments beneath such videos identified dozens of cases in which viewers describing themselves as elderly had altered actual medical treatment under the influence of synthetic advice.
One commenter, age 85, claims to have replaced omeprazole, a medication that suppresses stomach-acid production, with sweet potatoes.
A 77-year-old woman writes that she has not seen a doctor in three years and thanks a nonexistent physician for advice about dementia.
Gastroenterologists consulted for comment point to two simultaneous risks: a patient may postpone a real diagnosis for crucial weeks, or directly harm themselves by using an ineffective remedy or one incompatible with their current treatment.
Relatives often discover that treatment has been replaced only after the fact, when the bill comes not from the fraudulent channel but from the hospital to which the patient is admitted with complications caused by a delayed diagnosis.
The Scammer’s Main Weapon Is Not AI. It Is Human Fear
The technique used by the creators of these videos, a sharply alarming headline, a promise of imminent danger, and a demand for immediate action, has long been described in marketing textbooks as one of the most reliable ways to bypass rational fact-checking. A frightened viewer looks first for an immediate solution, not for the source of the information.
Older audiences are not necessarily more gullible in any general sense. But they are more likely to make such decisions alone, without an immediate opportunity to consult a relative or a doctor they know. It is precisely this circumstance that turns a momentary fright into an instant purchase rather than a question postponed for later consideration.
A Doctor’s Face Can Now Be Stolen in Minutes
Another dimension of the damage is the theft of real doctors’ identities.
At least five channels used the likeness of a well-known Brazilian gastroenterologist without permission to promote alarming health claims and natural alternatives to prescribed medications. The physician contacted YouTube and filed a police report in Sao Paulo.
Brazilian lawyers note that such activity formally falls under statutes covering fraud and identity theft. But enforcement consistently lags behind the scale of the phenomenon: while one investigation is underway, algorithms can generate dozens of new clones.
The Scammer Is in One Country, the Victim in Another, and the Money in a Third
Jurisdictional confusion compounds the problem at a systemic level.
A channel may be registered to an account in one country, targeted by advertising algorithms at an audience in another, and route its revenue through payment systems in a third. Neither Brazilian police, European regulators, nor YouTube itself has, on its own, a tool capable of shutting down the entire chain with a single decision.
International coordination mechanisms, from working groups of the Organisation for Economic Co-operation and Development on artificial intelligence regulation to bilateral memoranda between national consumer-protection agencies, are still discussing general principles rather than concrete emergency-response protocols for cross-border medical deception.
From Sao Paulo to Seoul: The Scheme Works the Same Way
The problem is by no means exclusively Brazilian or Portuguese-speaking.
South Korean regulators publicly acknowledged this summer that existing legislation is nearly powerless against artificial-intelligence channels posing as doctors. The law permits intervention only when a specific product is being directly advertised, and content creators methodically exploit that limitation.
One South Korean channel with more than 36,000 subscribers falsely claimed affiliation with a major university hospital and published videos with titles such as seven reasons why an internist with 25 years of experience drinks vinegar every day.
The structure of the deception, synthetic authority combined with a legal loophole, is reproduced from Sao Paulo to Seoul with almost no variation because the technology and the revenue model are the same everywhere.
Only the language of the subtitles and the nationality of the stolen face change.
Billions of Dollars Are Flowing to Scammers. And That Is Only the Visible Part
The broader context surrounding this phenomenon is staggering in scale.
According to the U.S. Federal Trade Commission, total fraud losses among Americans over the age of sixty nearly quadrupled in four years, rising from approximately $600 million in 2020 to $2.4 billion in 2024. The commission itself estimates that the actual damage, once widespread underreporting is taken into account, could be as high as $81.5 billion.
In its 2025 report, the Federal Bureau of Investigation recorded $7.748 billion in reported losses among the same age group and more than 201,000 victims, a 59 percent year-over-year increase.
Of that amount, the bureau directly attributes $352 million to fraud in which artificial intelligence played a decisive role.
Fake doctors on YouTube are only one of many tentacles of this industry. But this particular tentacle targets health rather than the victim’s wallet directly, making it potentially more lethal.
Federal agencies themselves acknowledge that the true scale of the problem is systematically understated. U.S. regulators estimate that fewer than one in twenty fraud victims reports the crime, meaning the billions in confirmed losses are merely the tip of an iceberg whose base remains statistically invisible.
YouTube Removes Channels. But Only After Millions of Views
Google, which owns YouTube, has limited itself to a formal response. The company emphasizes that its existing rules also apply to content created with artificial intelligence and prohibit medical misinformation capable of causing serious harm in the real world.
After researchers provided the company with the results of their review, the platform removed several of the largest channels in the sample, which together had accumulated approximately 41 million views.
The principle of preventing serious real-world harm sounds persuasive, but it remains difficult to apply as a tool of preventive control. The company reacts after the fact, in response to a specific complaint backed by evidence, rather than building a filter capable of stopping a new clone from being published before it accumulates its first hundred thousand views.
Since July 2025, the platform has formally tightened its policy on so-called inauthentic content. Videos dealing with sensitive subjects, including health, are required to display a prominent notice directly above the player indicating that the content is synthetic.
In early 2026, 16 major channels with a combined 4.7 billion views and approximately $10 million in annual advertising revenue were swept up in this crackdown, although most had been caught producing fake trailers and generic “AI slop” rather than medical misinformation specifically.
The case of the woman who believed the synthetic doctor is particularly revealing: the warning label was displayed above the video, and it did not stop her.
Labeling may reduce the platform’s legal exposure, but it does not necessarily protect an individual older viewer who grew up in an era when a white coat on a screen automatically signaled medical expertise.
Why Defensive Technology Is Already Losing to the Technology of Deception
The technological race between generating and detecting synthetic content is structurally tilted against defense.
The C2PA industry standard for content provenance, which YouTube and other platforms are gradually integrating into their moderation systems, cryptographically signs metadata recording how and with what tools a file was created. But it works only when the creator voluntarily agrees to labeling through a legitimate, registered generation tool.
For a scammer using unlicensed or deliberately modified models specifically to evade such signatures, the system presents no obstacle at all. The scammer simply does not embed metadata that is not mandatory at the point of upload anyway.
A detector that works after the fact by identifying visual and audio artifacts becomes outdated faster than it can be trained on each new generation of models. Commercial voice- and face-synthesis tools are updated every few months, while developing and validating a detection system takes entire quarters.
Europe Has Already Introduced Multimillion-Euro Fines. But There Is One Problem
The regulatory response is arriving late and with extreme inconsistency.
On August 2, 2026, less than a month ago, Article 50 of the European Union’s Artificial Intelligence Act took effect, for the first time requiring developers of generative systems to embed machine-readable labeling in synthetic content and requiring those who use the technology to create deepfakes to clearly disclose that fact to viewers.
Violators face fines of up to €15 million or 3 percent of global annual revenue, whichever amount is higher.
The European Commission is still finalizing the Code of Practice that will specify the technical standards for such labeling. The first draft appeared in December 2025, consultations concluded on June 3, 2026, and the final version is expected later this year.
In other words, the law is formally in force, while the practical tools needed to enforce it are still being built.
Brazil Became a Factory for the Problem While the Law Remained Stuck in Parliament
In Brazil, where most of the channels described above originated, the situation is the reverse: the law is lagging behind the technology more severely than almost anywhere else.
The proposed federal artificial intelligence law known as PL 2338 was unanimously approved by the Senate back in December 2024 and has been stalled in the Chamber of Deputies since March 2025.
In December 2025, the executive branch introduced an additional bill intended to resolve a constitutional conflict. The original text granted regulatory powers to the National Data Protection Authority in a manner that formally fell within the president’s exclusive constitutional authority.
Until that conflict is resolved, the final vote remains delayed, and the fines of up to 50 million Brazilian reais envisioned by the legislation remain hypothetical.
The gap between the country where the problem is being generated and the jurisdiction that first tightened the rules is not a technical detail. It is a structural reason why the synthetic-doctor industry was able to grow to its current scale in the first place.
Today a Fake Doctor. Tomorrow a Fake President
The technologies behind fake doctors, voice cloning, facial synthesis, and the generation of convincing speech, are fundamentally no different from tools that have already been used to interfere in electoral processes.
Two days before the New Hampshire primary in January 2024, thousands of voters received an automated call using a cloned version of President Biden’s voice urging them not to vote. Political consultant Steve Kramer, who organized the campaign, was fined $6 million by the Federal Communications Commission and faced 26 criminal charges.
The distance between a scammer selling sweet potatoes as a substitute for omeprazole and a political operative manipulating voter turnout is shorter than it appears. Both are shielded by the anonymity of generative models, and both rely on the calculation that an audience will trust a familiar voice faster than it will verify a fact.
In 2026, a commercial scam and an information operation use the same toolkit. The difference lies in the objective, not in the technology.
One Algorithm Can Deceive a Retiree and an Entire Country
The scale of the problem extends far beyond consumer fraud.
Analysts at the Atlantic Council’s Digital Forensic Research Lab noted as early as 2025 that hostile actors were systematically integrating synthetic media into established information-operation playbooks rather than inventing fundamentally new methods.
Between March and June 2025, OpenAI publicly disclosed and disrupted several China-linked networks, including an operation that generated thousands of fake social-media comments designed to simulate organic support for particular narratives.
During the military conflict between Iran and Israel in the summer of 2025, AI-generated videos depicting nonexistent missile strikes on Tel Aviv and supposedly downed F-35 fighter jets spread online in five languages within hours, while Israel’s own operation, code-named PRISONBREAK, began deploying synthetic content within an hour of actual strikes.
In its 2026 Global Risks Report, the World Economic Forum explicitly identifies disinformation as an amplifier of virtually every other systemic risk, from elections to financial crises.
The same technology that persuades an older woman in the morning to replace medication with sweet potatoes can, by evening, just as easily persuade an entire country that a nonexistent military catastrophe has occurred. The only difference is the size of the audience and who is paying to generate the video.
If Europe Closes the Door, Scammers Will Move Where the Law Is Weaker
The future trajectory is predictable in its basic logic, even if the details remain uncertain.
As fines under Article 50 begin to be applied in practice across the European Union, with the first high-profile enforcement cases likely within the next 12 to 18 months as the Code of Practice is finalized, creators of this type of content will shift their audiences toward jurisdictions where enforcement is weaker: Latin America, South Asia, and regions with fragmented regulatory systems.
The fact that instructional videos are already circulating in nine languages points directly toward such a strategy of regulatory arbitrage.
Brazil’s PL 2338, if the current pace is maintained, could reach a final vote in the Chamber of Deputies during 2026. But the history of its five-year journey through parliament provides ample reason for caution when predicting any timetable.
YouTube, for its part, will continue expanding automated deepfake detection and image-protection tools for individual doctors. But the technological race between detection and generation is structurally asymmetric: creating a convincing synthetic video is cheaper and faster than detecting and removing it.
The Most Reliable Defense Turned Out to Be Disturbingly Simple
At the individual level, the only filter that consistently works for now is not technological but behavioral: a viewer who discusses what they have seen with someone else before changing medical treatment is statistically far better protected than by any warning label above a video player.
That is why the woman at the center of this story, after realizing that she had been targeted by deception, directs her concern not at the regulator or the platform but at herself. She wants to learn how to distinguish synthetic content from authentic material.
It is simultaneously the most sober conclusion to the entire story and the most uncomfortable one. While governments and corporations argue over who should close the loophole first, the only practical line of defense remains the viewer sitting alone in front of a screen.
An Algorithm Will Always Find a Patient
The gap between the letter of the law and the experience of an individual viewer is the central conclusion of the entire story.
Brussels can adopt as many provisions as it wants concerning machine-readable labeling and fines calculated as a percentage of corporate revenue. A law takes effect in lawyers’ offices long before it takes effect in the mind of an 82-year-old woman scrolling through YouTube in the evening in search of advice on how to preserve her eyesight.
The synthetic-doctor industry will not disappear because of a single directive or a single investigation. It is built according to the same logic as advertising itself, except that one final restraint has been removed: there is no longer a real person who can be held accountable for what was said.
As long as this gap remains open between the technology that manufactures trust and the technology meant to verify it, an algorithm will always find a patient willing to believe, and a channel will always be ready to profit from that belief.