If You Want to Sell It, Show It

If the goal is attention and profit, how do you account for everything that cannot be measured when generating an ad?”

Isabelle Holloway

How many times can you reproduce something before it's no longer considered the original?

Take Marilyn Monroe. First, there is Monroe herself. Then there is Monroe made up in a white satin dress for a premiere, a paparazzi photograph in New York City, a movie poster for Niagara, Andy Warhol’s silkscreen prints, an AI-generated Monroe who never existed, someone dressed as Monroe, or a platinum bob wig on a mannequin.

At what point do we stop looking at Monroe and start looking at an idea of Monroe?

Now replace Monroe with America. The United States has always existed somewhere between a nation and an idea of a nation. The Revolution became a mythology in which the “Minute Man” traded his plow for a musket to stand against an empire and spark a new political experiment. This experiment drew immigrants across deserts and oceans with hopes of a better life, while later generations tried to turn those hopes into reality.

Yet ideals of liberty and equality have always existed alongside contradictions like slavery, the dispossession of Native peoples, and the exclusion of women and others from power. The tension between dream and reality makes America into a space of projection. Each generation can project its own hopes and decide what America has been, what it is, and what it could become.

The Liberty Tree, rattlesnake flag, Conestoga wagon, log cabin, steam locomotive, assembly line, white-picket-fence suburbia, and moon landing became symbols of American identity. Today, we might add delivery boxes, self-landing rockets, smartphones, and satellite constellations. None of these things are America, but we recognize them as America. We turn reality into symbols and myths because they make it easier to understand. Eventually, those symbols become part of our reality.

AI is extraordinarily effective at turning the symbolic into convincing representations of reality. If you ask it to generate a diner, it assembles the familiar symbols: a ‘googie’ exterior, checkered floors, red booths, milkshakes, a waitress, and a jukebox. The result may look like a diner that does not exist.

I am not anti-AI. Even though its development often seems to outpace legislation and consideration for its effects on the job market, AI is an amazing tool with enormous potential. It already improves medicine, engineering, and safety systems; translates languages, personalizes education, and creates visual effects otherwise impossible. What I am skeptical of is AI-generated advertising.

AI-generated ads are often too cheerful, generic, and perfect. Many customers instinctively assume an AI ad is spam. More importantly, they can advertise products or experiences that never existed. A bakery can generate a slice of pie that has never been served. A motorcycle rally can show bikers who are not your friends and neighbors. A town can look like an uncanny version of itself without its actual layout or character. Whether intentional or not, the message can feel made without much care.

In a small town, the distinction between artificial and real matters. You may know the line cook and have actually eaten the food. A photograph of a repair shop tells you who works there, while a grease-smudged wall, stacked boxes, or duct-taped sign tells you it is lived in.

Jean Baudrillard developed the idea of the simulacrum, a representation detached from the reality it originally represented. AI can create convincing representations with no reality behind them. A photograph can misrepresent something that exists; AI can produce the appearance of something that never did.

Consider Atlantic City: an AI-generated souvenir postcard might combine pristine sand, turquoise water, neon casino signs, and Gilded Age boardwalk glamour. But the Atlantic City I remember has gray, foamy tides, driftwood and horseshoe crabs, sand in my flip-flops, the steel-skeleton ghosts of roller-coasters like the Loop the Loop, and cheap arcades and dive bars filled with awkward teenagers. The unmanicured portrait feels more alive because the postcard prioritizes the idea of a perfect beach vacation over the charms of reality.

There is a complication with old postcards: they were also carefully staged and mass-produced. Yet, time changes what reproduction means. Chemical changes and the distance of the past give the image a history that a new reproduction cannot possess. Walter Benjamin called this an object's “aura,” tying its image to its unique existence in the “here and now.”

A generated image gives us perfect representation without history. The problem is not that it will never yellow or fade, but that it is not anchored to anything in this world.

Then there is abundance. In minutes, a bar can generate dozens of happy-hour, holiday, and event graphics. But when everyone can produce unlimited images using the same software, visual culture becomes hostile in its uniformity. If everything can look like a statistical ideal of perfection, then nothing is particularly meaningful.

Then I wonder: what happens when the tool becomes the advertiser?

A person uses AI to make an advertisement. AI draws from human culture, produces an image, tests its engagement, generates another version, and tests again. Eventually, the system can optimize toward clicks, sales, or attention while requiring less human intervention.

An algorithm optimizing for engagement does not understand authenticity. AI can get the response we want without understanding why that response matters.

This is where the slightly terrifying paperclip maximizer thought experiment comes in: a system given one objective may pursue it without understanding everything humans value around it.

AI lacks human spontaneity, intuition, and lived experience. It can reproduce the appearance of a finished work, but not the life that made the work necessary. Nikola Tesla’s inventions came from his particular vision of the future; Sergei Rachmaninoff’s symphonies carried the homesickness of his exile. Their work was not a calculation of choices, but the residue of their particular lives and experiences.

Advertising has the same problem: if the goal is attention and profit, how do you account for everything that cannot be measured when generating an ad?

Edgar Allan Poe wrote that “There is no exquisite beauty… without some strangeness in the proportion.” A person who does not look like a runway model looks familiar; a cluttered kitchen feels homey. Photographers look for what is out of place because specificity gives people something meaningful to admire. Often, imperfections are not incidental, but the part we remember most.

During the Revolution, the rattlesnake became a symbol of American unity. A later version took the form of an ouroboros, a snake eating its own tail.

AI can create a similar loop, consuming human visual culture and producing new images before those images enter the cultural archive. People become accustomed to those images, businesses imitate them, and artists reproduce the aesthetic. AI learns from those reproductions and generates the next version.

The snake eats its own tail as the representation becomes material for the next representation.

What are we actually selling when we use AI to generate advertisements? If you want to sell it, show it. Show us the person who actually works there, even if they have some stubble or are caught in the middle of a hard day’s work. Show us the motel with a letter falling off its sign. Show us the hot dog that is slightly burnt but that I would still happily eat. Show us the cluttered convenience store where you know where everything is and the owner knows you. Show us the America that exists.

In an age when AI can generate almost anything, we should value what no machine will ever be able to manufacture: the life behind the person, the history behind the postcard, the reality behind the advertisement, and the human spirit behind it all.