The screen glowed in the dark living room, casting a neon pink hue across the floor. I was scrolling through social media, half-asleep, when the video started playing. It didn't look like a public service announcement. It didn't look like the gritty, terrifying anti-drug ads of my childhood—the ones with frying pans, smashed windshields, and weeping parents.
Instead, this video looked like a dream.
Swirling colors melted into each other. Shimmering, iridescent dust floated around a hyper-stylized digital model whose face radiated pure euphoria. The music was a hypnotic, lo-fi beat that thrummed right through my chest. For a split second, my brain didn’t process the warning text flashing at the bottom of the screen. My brain just registered a feeling.
Warmth. Escape. Beauty.
Suddenly, I caught myself. I knew what the video was trying to do. It was part of a major government campaign aimed at warning teenagers about the dangers of synthetic opioids like fentanyl. Yet, here I was, a grown adult, staring at the screen thinking, That looks incredible. If a well-intentioned warning can accidentally double as a luxury perfume ad for narcotics, we have crossed into dangerous new territory.
The campaign in question was built using generative artificial intelligence. The creators wanted to reach Gen Z on platforms like TikTok and Instagram, choosing a medium they thought would resonate with digital natives. They wanted to speak the language of the internet. They succeeded in capturing attention, but they failed fundamentally at human psychology.
When you use AI to generate imagery of substance use, the technology strips away the grime. It eliminates the friction of reality. AI excels at making things look polished, ethereal, and hyper-real. It creates a flawless aesthetic out of chaos.
But drug addiction is not an aesthetic.
The Hallucination of the Perfect High
To understand why this happens, look under the hood of how these visual models actually work. An AI image generator does not understand what addiction is. It does not know about the sweat, the shaking, the desperate calls to a dealer, or the quiet grief of a family left behind. It only understands pixels, patterns, and prompts.
When an engineer types a prompt like "visual representation of a psychological high," the machine combs through millions of scraped internet images associated with bliss, euphoria, and altered states. It pulls the vibrant colors of festival lights, the soft focus of dream sequences, and the flawless skin of digital avatars. It blends them into a seamless, mesmerizing loop.
The machine sanitizes the tragedy.
Consider a hypothetical teenager named Leo. He is sixteen, sitting in his bedroom, feeling the crushing weight of academic pressure and social isolation. His anxiety feels heavy, sharp, and intensely real. He opens his phone and sees an AI-generated anti-drug ad.
The ad is meant to show the danger of escaping reality. But because the AI has rendered the "escape" as a gorgeous, swirling nebula of pastel light, Leo doesn’t see a warning. He sees a contrast. He compares his messy, painful, concrete reality with the smooth, beautiful digital illusion on his screen.
The ad tells his intellect to stay away, but it tells his visual cortex that relief looks beautiful.
This is the boomerang effect in public health communication. When a message is too stylized or too abstract, the brain rejects the intended moral and absorbs the aesthetic instead. We saw flashes of this in the late 1990s with traditional media, but AI accelerates the problem exponentially. It allows for the rapid, cheap production of infinite content that appeals directly to the subconscious mind, unmoored from human empathy.
The Invisible Stakes of Automated Empathy
Human graphic designers and directors know how to walk this tightrope because they possess lived experience. A human director knows that to convey the horror of fentanyl, you don't make the user look like a cyberpunk deity. You focus on the trembling hand, the cold sweat, the empty room. You ground the narrative in the physical world because that is where the suffering happens.
When we outsource our cautionary tales to algorithms, we lose control of the tone.
The creators of the campaign defended the work by pointing to engagement metrics. The videos gained millions of views. Comments poured in. On paper, the campaign looked like a massive success to a data analyst sitting in an office.
Data, however, is a terrible measure of human behavior.
A click is not a conversion to sobriety. A view is not a saved life. People stop to watch train wrecks, and they stop to watch beautiful, bizarre AI animations. If millions of teenagers watch an ad because it looks cool, the metric is high, but the efficacy is zero. Or worse, negative.
The real danger of synthetic drugs like fentanyl is their invisibility. They don't look like poison; they look like ordinary prescription pills. They are cheap, ubiquitous, and lethal in microscopic doses. The fight against them requires absolute clarity. It requires pulling back the curtain on how deceitful these substances are.
By wrapping the warning in an AI-generated fantasy, the campaign accidentally mirrored the exact lie that drug dealers tell: that this substance will make your world more beautiful.
Shifting the Lens Back to Reality
We have to ask ourselves what we want from our public health messages. Do we want them to look like high-end music videos, or do we want them to save lives?
True persuasion doesn't come from a perfectly rendered algorithm. It comes from the jagged, unpolished truth. It comes from the voice of a mother who lost her child, the raw testimony of a survivor, or the unvarnished reality of a recovery ward. Those things are uncomfortable to look at. They don't perform well as background aesthetics for casual scrolling. But they stick to the ribs. They linger in the mind when temptation strikes.
The solution isn't to ban AI from public health entirely. The technology can be used to map distribution data, analyze search trends to deploy resources, or translate educational materials into dozens of languages instantly.
But when it comes to telling the story of human suffering and human survival, the machine must step aside.
The pink glow on my living room floor eventually faded as the video looped and ended. I shut off the screen, leaving the room in complete, dark silence. In that quiet, the memory of the glittering digital dust felt incredibly hollow. It was a ghost of an experience, manufactured by code, designed to warn me but engineered to entice.
We must ensure our warnings sound like a human voice shouting a frantic, desperate alert from the edge of a cliff, not a synthesized lullaby humming us to sleep on the way down.
