COLUMN: Flirting bananas, crying strawberries—welcome to TikTok’s algorithmic gutter

How to Make a Bananito in 47 Seconds

Here’s what it takes to create a dramatic AI fruit video in 2026: a computer, access to an AI image generator, a text-to-speech synthesizer, and about 47 seconds of human labor. The rest—the script, the animation, the facial expressions, the emotional soundtrack—is fully automated.

The production cost? Virtually zero. The potential for ad revenue from hundreds of millions of views? Colossal. The equation is so lopsided that it makes any form of human creation economically irrational in the eyes of the platforms.

An illustrator spends eight hours drawing a character. A screenwriter spends weeks writing an episode of a series. A human TikTok creator spends hours filming, editing, and fine-tuning a video. And all that to garner fewer views than a banana generated in less than a minute.

The “slop” economy: who wins, who loses

The accounts that post these AI-generated fruit videos aren’t artists. Most of the time, they aren’t even identifiable individuals. They are content farms—industrial operations, often based in countries where the cost of digital labor is negligible, that produce hundreds of videos a day and flood all platforms with them simultaneously.

Their business model is simple: volume, volume, volume. Out of every thousand videos posted, just one needs to go viral to make the entire operation profitable. And when the algorithm systematically rewards this type of content, the odds become even more favorable.

Who loses out in this equation? Everyone. Human creators, whose work is drowned out by a tsunami of automated drivel. Viewers, whose attention is captured by content designed to stimulate without nourishing. And the platforms themselves, which are sawing off the branch they’re sitting on by degrading the quality of the user experience at a dizzying pace.

When Dopamine Replaces Meaning

There’s a term in neuroscience to describe what AI fruits do to your brain: dopaminergic hijacking. Every plot twist—will Bananito choose strawberry or mango?—triggers a micro-release of dopamine, the molecule of pleasure and anticipation that helped our ancestors survive by driving them to explore, search, and want to know what comes next.

The problem is that this mechanism was designed to help us find food and avoid predators—not to keep us glued to a love story between virtual citrus fruits at 3 a.m.

Dr. Judson Brewer, a neuroscientist and research director at Brown University, has documented how digital reward loops exploit the same circuits as addictive substances. The only difference is that the dealer is an algorithm, the substance is content, and the syringe is a 6.7-inch screen you carry in your pocket.

Children on the Front Lines

Here’s the part TikTok executives would rather you didn’t read. The main audience for AI fruit videos isn’t jaded adults scrolling out of boredom during their lunch break. It’s children. Preteens. Developing brains whose critical-thinking circuits aren’t yet fully formed.

A nine-year-old doesn’t watch Bananito with the ironic detachment of an adult who knows it’s algorithmic drivel. He watches Bananito just as he would watch a cartoon. They become attached. They take sides. They comment. They come back the next day to see what happens next. And with every viewing, their brain strengthens the neural connections that associate this type of passive stimulation with pleasure.

We’re wiring an entire generation to prefer machine-generated emotional drivel over any form of authentic human storytelling. And we’re doing it in broad daylight, in full view of everyone, calling it a fun trend.

Reality TV: The Final Frontier for AI

It’s no coincidence that AI systems have adopted the romantic reality TV format. The genre is, by its very nature, the narrative form easiest for artificial intelligence to simulate. Why? Because reality TV relies on a limited number of emotional patterns—jealousy, betrayal, reconciliation, love at first sight, breakups—that repeat endlessly according to predictable formulas.

A language model doesn’t even need to be sophisticated to generate this type of scenario. All it needs is to have “ingested” a few seasons of Love Island, Les Marseillais, or The Bachelor to produce endless variations on the same archetypes: the seducer, the rival, the heartbroken character, and the surprise return.

The fact that these archetypes are embodied by fruits rather than humans doesn’t diminish the formula’s effectiveness. It’s actually an advantage: the absurdity of the visuals creates an extra layer of engagement. We watch because it’s ridiculous. We stay because it’s addictive.

The Silent Death of the Screenwriter

While Bananito racks up millions of views, human screenwriters—people who have studied their craft for years, who master the subtleties of dramatic structure, who know how to create characters that stay with us for decades—can no longer find work.

The 2023 Hollywood writers’ strike was already about this very issue. Three years later, the answer has arrived, and it’s even more brutal than the strikers feared. AI hasn’t replaced screenwriters in the studios. It has created an entire parallel universe where screenwriters aren’t even necessary—a universe that attracts more attention than anything Hollywood produces.

And yet, no one is protesting outside TikTok’s offices.

Follow the money, not the fruit

Every AI fruit video that racks up millions of views generates ad revenue. That revenue fuels content farms that produce more videos. Those videos generate even more views. The cycle continues. The snake bites its own tail, and with every turn, it grows bigger.

But the real money isn’t in the ads that play before a video of a love-struck banana. The real money is in the data.

Every second you spend watching Bananito is a second of behavioral data that TikTok collects, analyzes, and monetizes. Your watch time. Your reactions. The exact moment you almost scrolled past but ultimately kept watching. Everything is recorded, everything is exploitable, everything has market value.

AI fruits aren’t a product. You are the product. AI fruits are just the most effective trap ever devised to keep you glued to the screen while your data is being extracted.

TikTok knows. TikTok does nothing.

TikTok has the tools to detect AI-generated content. The platform has sophisticated detection systems, moderation teams, and official policies on synthetic content. And yet, AI fruit videos proliferate without the slightest friction.

Why? Because content that generates hundreds of millions of views is content that generates revenue. And in the attention economy, revenue comes first—before quality, before users’ mental health, before the truth itself.

TikTok’s moderation system is designed to remove dangerous content. But a banana flirting with a strawberry isn’t dangerous in the traditional sense. It doesn’t violate any laws. It doesn’t spread factual misinformation. It doesn’t depict any violence. It does something far worse: it normalizes the idea that content doesn’t need to be human to deserve your attention.

What Our Clicks Say About Us

It would be easy to blame the algorithm. To point the finger at content farms. To denounce TikTok. And all of that would be justified. But there’s a more uncomfortable question that no one wants to ask: why does it work?

Why do tens of millions of human beings—endowed with consciousness, free will, and unlimited access to all the knowledge ever produced by humanity—choose to spend their time watching artificial fruits act out prefabricated romantic dramas?

The answer is unsettling. AI-generated content doesn’t create a need—it fills a void. A void of social connection. A void of meaning. A narrative void in lives where the complexity of reality has become so overwhelming that the simplicity of a drama involving citrus fruits offers an accessible emotional refuge that requires little cognitive effort and is immediately gratifying.

The Attention Cliff

Tim Wu, a researcher in the economics of attention and author of The Attention Merchants, predicted as early as 2016 that the final frontier of commercial exploitation would be neither land, nor oil, nor even data. It would be human attention itself—that finite, non-renewable resource, of which each of us has about 16 hours per waking day.

AI-generated content represents the next stage of this colonization. Content so easy to consume, so devoid of intellectual friction, that it captures your attention without you even realizing it. You don’t decide to watch Bananito. You just find yourself watching it. And when you look up, twenty minutes have vanished.

Twenty minutes you didn’t spend reading. Thinking. Creating. Talking to a real person.

The Frog-in-Lukewarm-Water Syndrome, Digital Edition

Five years ago, if you’d told someone that the most-watched videos on the world’s largest entertainment platform would be AI-generated characters in romantic situations, they would have laughed in your face. They would have told you it was a satirical dystopia, an episode of Black Mirror too far-fetched to be believable.

And yet, here we are. And no one is laughing.

This is precisely how normalization works. Each step, taken in isolation, seems harmless. Fun filters. Celebrity deepfakes. AI avatars. Talking animals. Fruits hitting on people. Each step taken makes the next one acceptable. And when you look up to gauge how far you’ve come, you realize you’re very, very far from where you started.

The question is no longer: How did we get here? The question is: Where does it end?

The historical precedent no one mentions

In the 1950s, American television went through a similar crisis. Rigged game shows—Twenty-One, The $64,000 Question—captivated tens of millions of viewers. The ratings were astronomical. Advertisers were thrilled. And everyone knew, more or less vaguely, that something wasn’t right.

When the scandal broke and the public discovered that the results were rigged, trust in television suffered damage that some historians believe has never been fully repaired.

TikTok’s AI-generated content is the rigged quiz show of our time. The difference is that this time, everyone knows it’s rigged—and nobody cares.

The Numbers Behind the Flood

According to the most conservative estimates, AI-generated content now accounts for between 15% and 25% of everything posted daily on major social media platforms. On TikTok specifically, some analysts estimate that this percentage exceeds 30% when videos partially assisted by AI are included.

At this rate, the majority of content visible on social media will be synthetic by the end of the decade. This isn’t an alarmist prediction—it’s a linear projection based on current growth rates.

For human creators, this means one very concrete thing: they are no longer competing with other humans. They are competing with machines capable of producing in one minute what takes them hours, at an infinitely lower cost, and often with a higher rate of virality.

The Tragedy of the Artisan Facing the Factory

Imagine a cabinetmaker who spends six months carving a solid-wood table. Every detail is carefully considered. Every joint is perfect. The result is a work of art. Then one day, a factory opens across the street from his home and begins producing plastic tables that, from a distance, resemble wooden tables—for one-hundredth of the price.

Most customers choose the plastic table. Not because they think it’s better—they know it isn’t—but because, in their eyes, the difference in quality doesn’t justify the difference in price.

That’s exactly what’s happening on TikTok. And this time, the plastic table is a banana with a certain charm.

Solutions that exist but are too expensive

The technology to systematically label AI-generated content exists. Digital watermarks—invisible signatures embedded in every image or video produced by an AI generator—are technically mature. Companies like Google, OpenAI, and Adobe have developed standards. The C2PA consortium offers a comprehensive authentication framework.

But none of these solutions is mandatory. No law requires platforms to implement them systematically. And no platform has an economic incentive to do so, since labeling AI-generated content would reduce engagement—and thus revenue.

This is the fundamental paradox: transparency tools exist, but for platforms, using them would be like sawing off the branch they’re sitting on. And in a market where every percentage point of engagement translates into billions of dollars in market capitalization, no one wants to be the first to start sawing.

The regulatory vacuum is a political choice

The European Union, with its AI Act—which has been phased in since 2024—has laid the groundwork for a regulatory framework. But the process is slow, exemptions are numerous, and enforcement remains in its infancy. In the United States, the gap is even wider: no federal legislation specifically regulates synthetic content on social media.

And yet, if a food manufacturer sold products without labels, without a list of ingredients, and without disclosing the presence of additives, it would be sued within 24 hours. Why do we accept this for “cognitive food” that we would categorically reject for physical food?

Because technology lobbies are more powerful than food lobbies. Because the elected officials who should be enacting legislation are themselves dependent on these platforms for their political communication. And because the public, numbed by constant entertainment, has not yet realized that this issue directly affects them.

What do we lose when we can no longer distinguish between what is real and what is fabricated?

There is a famous psychological experiment, conducted in the 1970s by researcher Stanley Milgram, in which participants agreed to administer what they believed to be painful electric shocks to other people, simply because an authority figure asked them to. The experiment demonstrated the terrifying ease with which human beings relinquish their judgment when faced with an institutional framework.

Recommendation algorithms are the new authority figure. They don’t wear white coats. They don’t give explicit orders. But they shape, second by second, what you see, what you feel, and gradually, what you think. And unlike Milgram’s experimenter, they have no moral conscience.

When an entire generation grows up in an environment where the line between human-generated and synthetic content is invisible, it is not just the notion of authenticity that crumbles. It is the very ability to assess reality that erodes.

The Reverse Turing Test

For a long time, the question has been: Can a machine think like a human? That was the Turing test. The relevant question in 2026 is exactly the opposite: Can a human still distinguish what is human from what is not?

And if the answer is no—if hundreds of millions of people watch synthetic content knowing it’s synthetic and not caring—then it’s not the machine that passed the test. It’s the human who failed it.

Bananito will never pass the Turing test. It doesn’t need to. It has achieved something far more disturbing: it has proven that the test no longer matters.

No magic formula—but acts of resistance

The first thing to do is to name the problem. AI slop. Algorithmic mush. Call it what you will, but stop calling it a trend. A trend is a passing fad. What’s happening on TikTok is a structural shift in the way content is produced, distributed, and consumed. It’s a transformation, not a fad.

The second thing is to protect children. Not with rhetoric. With laws. Laws that require synthetic content to be labeled, laws that limit minors’ algorithmic exposure, laws that hold platforms accountable for what they amplify.

The third thing is to support human creators. Not out of charity—but for the sake of cultural survival. Every time you share the work of a human artist, musician, writer, or videographer, you’re performing an act of resistance against the machine. A small act. But small acts, when added together, build dams.

The choice that’s still yours

Your attention is the last thing you possess that algorithms cannot take from you by force. They can seduce you. They can manipulate you. They can make resistance uncomfortable. But the thumb that scrolls is yours.

The next time the algorithm shows you a crying piece of fruit, you’ll have exactly two seconds to decide. Two seconds to choose between the numbing comfort of mindless content and the effort—sometimes uncomfortable, often demanding, always necessary—of seeking out something true.

Two seconds. That’s all that separates a passive spectator from an engaged citizen.

The Most Honest Fruit in the World

Paradoxically, Bananito may be the most accurate reflection of our digital condition. It’s empty—and yet we still look at it. It’s manufactured—and yet we still grow attached to it. It’s disposable—and yet we still crave more of it.

And yet, somewhere amid this absurd frenzy, there is a truth that neither the algorithm nor the machine has been willing to reveal. That truth is that the human appetite for stories is so deep, so fundamental, so ingrained in our neurological wiring, that it can be satisfied even by the poorest of imitations.

This is both our greatest vulnerability and our greatest hope. Vulnerability, because this appetite makes us susceptible to manipulation. Hope, because it means that the day we collectively decide to feed this appetite with real stories, real faces, and real voices—on that day, the algorithm’s plastic fruits will return to the digital void from which they should never have emerged.

The Final Word

In the meantime, Bananito will continue to flirt, cry, betray, and make up—before an audience of hundreds of millions of pairs of eyes that have forgotten that looking is not the same as seeing.

And somewhere, in a ByteDance office, someone will watch the engagement numbers climb and smile. Because in the attention economy, the most profitable content is the kind that costs the least to produce and holds viewers’ attention the longest.

A banana with eyes. Zero cents’ worth of human creativity. Hundreds of millions of views.

If that isn’t a sign of a civilization with a problem, then the word “problem” has lost all meaning.

Signed, Jacques PJ Provost