Deepfakes: AI hoax fooled 20% of viewers
A 20-second AI-generated video managed to fool 20% of viewers. The joke reveals how fragile the truth has become in the deepfake era.
English · Original discussion in Spanish · Published
The deepfake that got through: 20% of viewers failed to spot the hoax
A video generated by artificial intelligence, barely 20 seconds long, managed to convince one in five viewers that it was real. The figure, drawn from a spontaneous reaction on social media, is no anecdote: it reflects the growing difficulty of telling the real from the synthetic in an era where AI is blurring the lines of truth.
The video, presented as a supposed piece of investigative journalism, included explicit references to online communities with very distinctive profiles. Those who watched it called it "brutal" and "brilliant", but also issued a warning: most people didn't catch on until a detail — the pronunciation of acronyms like "es-u-uve" instead of SUV — gave the game away. Even so, according to estimates, around 20% of the first viewers actually believed it was authentic.
The line between humour and disinformation
The video operated in the realm of parody, but the danger lies not in the comedic intent but in the technical quality. Generative AI can now imitate voices, faces and contexts with a fidelity that a few years ago required professional studios. The fact that a homemade video, probably the work of an amateur with accessible tools, managed to fool a significant fraction of its audience is a warning sign.
Some analyses suggested the believability lay in the mix of recognisable references with a news-broadcast format. "When the format is familiar, the brain tends to accept it without filtering," note experts in cognitive psychology. The problem gets worse when the content appeals to ideological biases or group affinity: the very people it targets become the first to share it without verifying it.
The cost of credulity in the AI era
The segarro anecdote reveals a structural phenomenon. The same technology that makes deepfakes to laugh about among friends can be used to manipulate elections, crash stock prices or discredit public figures. The 20% figure is not an official statistic — it emerged from a comment in the reaction thread itself — but it matches academic studies that put the believability rate of medium-quality deepfakes at between 15% and 30%.
Those who champion digital education as the antidote point out that critical-thinking training has been deficient for decades. "The new illiterates will be those who cannot tell the real from the manipulated," one of the comments warned. The difference is that technology is now advancing faster than education programmes.
How can you protect yourself?
The warning signs remain the same: verify the original source, look for inconsistencies in lip movement, shadows or reflections, and distrust any content that triggers an intense emotional response. But the margin of error is shrinking. The video in question only gave itself away through a tiny verbal slip; without that mistake, the 20% would have been far higher.
Given all this, the question is not whether someone will fall for a deepfake, but when you will. And what is more unsettling: would you be able to spot it?
Most shared
Related discussions on the forum
A video generated by artificial intelligence, barely 20 seconds long, managed to convince one in five viewers that it was real. The figure, drawn from a spontaneous reaction on social media, is no anecdote: it reflects the growing difficulty of telling the real from the synthetic in an era where AI is blurring the lines of truth.
The video, presented as a supposed piece of investigative journalism, included explicit references to online communities with very distinctive profiles. Those who watched it called it "brutal" and "brilliant", but also issued a warning: most people didn't catch on until a detail — the pronunciation of acronyms like "es-u-uve" instead of SUV — gave the game away. Even so, according to estimates, around 20% of the first viewers actually believed it was authentic.
The line between humour and disinformation
The video operated in the realm of parody, but the danger lies not in the comedic intent but in the technical quality. Generative AI can now imitate voices, faces and contexts with a fidelity that a few years ago required professional studios. The fact that a homemade video, probably the work of an amateur with accessible tools, managed to fool a significant fraction of its audience is a warning sign.
Some analyses suggested the believability lay in the mix of recognisable references with a news-broadcast format. "When the format is familiar, the brain tends to accept it without filtering," note experts in cognitive psychology. The problem gets worse when the content appeals to ideological biases or group affinity: the very people it targets become the first to share it without verifying it.
The cost of credulity in the AI era
The segarro anecdote reveals a structural phenomenon. The same technology that makes deepfakes to laugh about among friends can be used to manipulate elections, crash stock prices or discredit public figures. The 20% figure is not an official statistic — it emerged from a comment in the reaction thread itself — but it matches academic studies that put the believability rate of medium-quality deepfakes at between 15% and 30%.
Those who champion digital education as the antidote point out that critical-thinking training has been deficient for decades. "The new illiterates will be those who cannot tell the real from the manipulated," one of the comments warned. The difference is that technology is now advancing faster than education programmes.
How can you protect yourself?
The warning signs remain the same: verify the original source, look for inconsistencies in lip movement, shadows or reflections, and distrust any content that triggers an intense emotional response. But the margin of error is shrinking. The video in question only gave itself away through a tiny verbal slip; without that mistake, the 20% would have been far higher.
Given all this, the question is not whether someone will fall for a deepfake, but when you will. And what is more unsettling: would you be able to spot it?
Most shared
Related discussions on the forum
- The sale of commercial software is dead · 192 replies
- The forum is DYING. The lack of users and content is starting to be worrying. · 210 replies
- The PSOE (Spain's Socialist Party) placed Sara Fruta on €50,000 a month · 56 replies
- Is it known yet who the 61 hitman journalists working for the Cafre's sewer were who lynched the rest of the journ · 27 replies
Also available in: Deutsch
Summary of a discussion on Burbuja.info - Foro de economía, actualidad y política., translated from Spanish and reviewed before publication. Read the full discussion (100 replies).
More summaries
-
Spain announces veto on social networks for minors under 16
Spain prepares a ban on social media for minors under 16 while Brussels drafts its EU Kids Act. The challenge: verifying age without identifying everyone.
-
Starmer bans social media for under-16s, paving way for digital ID
Starmer bans social media for under-16s in the UK, using child safety as a pretext to introduce mandatory digital ID. Elon Musk warns of a surveillance state.
-
Kirkaversary: Viral Social Media Trend Honoring Charlie Kirk's Death Anniversary
The 'Kirkaversary,' a viral celebration of Charlie Kirk's death anniversary, sparks debate online. Understand the phenomenon.
-
Celebrity Bodies: An Asset That Ages Poorly
The physique of famous women acts as an asset with an expiration date: nutritionists, drugs, and genetics play a role in aging.
-
Putin and Russia accused of using AI to fake images of riots in France
BFM Business accuses Putin of spreading fake images of the October 17 looting. AI as a destabilisation weapon. Key facts about a growing phenomenon.
-
Dugin warns of what is happening in France
Russian ideologue Alexandr Dugin says on X that France is "imploding from within" and calls for a vote for Le Pen. His post comes after riots in several French cities.