There is a thriving genre of internet video where a person holds two nearly identical lipsticks up to the camera and asks you to guess which one is real. Same gold tube. Same logo. Same satisfying click. One cost forty dollars at a counter, the other cost six dollars from a marketplace seller with a five-star rating and no return policy. The reveal is always a small thrill, and it is almost always a coin flip. You cannot tell. Neither can the camera.
That uncertainty is the whole business model of the counterfeit beauty industry, and it is worth a teacher's attention. Not because educators need a new reason to worry about lip gloss, but because the question at the center of the dupe economy is exactly the question we keep asking about student writing in the age of AI. How do you prove something is real when the fake looks identical from the outside?
The dupe economy runs on the gap between looking real and being real
Counterfeit cosmetics are a multibillion dollar market, and they have gotten very good. The packaging is convincing. The shade names match. The barcodes scan. What you cannot see is what is inside the tube, and that is where the story turns from a shopping cautionary tale into something more serious. Investigations into seized counterfeit makeup have turned up lead, arsenic, mercury, and in some memorable cases, ingredients you would more readily associate with a bathroom than a bathroom cabinet.
The consumer's problem is not that they are careless. The problem is that they are being asked to verify authenticity using only the surface. The label says what the seller wants it to say. The tube performs the part. And the actual evidence, the supply chain, the lab testing, the chain of custody from manufacture to shelf, is invisible at the moment of purchase. By the time the harm shows up, on your skin or in a lab report, the transaction is long over.
Ingredient transparency was supposed to fix this. In theory, a label is a promise. In practice, "fragrance" can legally hide dozens of undisclosed compounds, "natural" means almost nothing, and a "clean beauty" claim is a marketing posture, not a regulated standard. The most honest thing on many beauty products is the price, and even that lies.
Surface checks fail in predictable ways
Here is the part that should sound familiar to anyone who has sat through a faculty meeting about AI. The instinct, when fakes get good, is to get better at spotting fakes. Learn the font weight on the authentic box. Memorize the texture of the genuine cap. Run the serial number through an app. Squint harder.
This is forensic authentication, and it has a ceiling. Every tell you learn, the counterfeiter learns too, usually faster, because they have a financial incentive and a feedback loop you do not. The mismatched font gets fixed. The hologram gets copied. The app gets spoofed. You are in an arms race where the other side iterates for a living, and you are doing it in the checkout line with your phone at three percent battery.
We are running the same race in education right now. A student turns in an essay. It reads cleanly. Did they write it, or did a model? So we reach for the cosmetic equivalent of squinting at the box: AI detectors that score a paragraph and announce a probability. The trouble is that the detector is also just looking at the surface, the finished text, and trying to reverse-engineer its origin. It cannot see the supply chain either. It is guessing from the tube.
The detector is a counterfeit test that catches the wrong people
The deeper problem with after-the-fact detection, whether for makeup or for prose, is not just that it can be beaten. It is that it generates confident errors, and those errors land on real people.
A forensic AI detector flags a student. The paragraph "scored high." But these tools are notoriously unreliable, and they fail in a particular, unfair direction. Writing that is plain, formulaic, or grammatically tidy reads as machine-made to a detector, which means English language learners and students who write in a careful, rule-following register get flagged more often. They wrote the thing themselves. The test said otherwise. That is not a verification system. That is a counterfeit test that occasionally accuses the genuine article of being fake.
Imagine if the lipstick scanner returned a false "counterfeit" verdict on one in ten authentic products, and disproportionately on products from certain regions. We would call that scanner broken. We should hold a tool that adjudicates a student's integrity to at least that standard.
Authenticity you can only guess at after the fact is not authenticity you can prove. It is a vibe with a confidence score attached.
Transparency beats forensics, in the aisle and in the assignment
So what actually works? In the beauty world, the answer is not a better fake-spotting trick. It is provenance. Buy from the authorized retailer. Trace the product to its source. The brands and shoppers who care about authenticity are moving toward verifiable supply chains, batch codes that actually resolve, sourcing you can audit. The shift is away from "does this look real" and toward "can you show me where this came from."
That is the right move for student writing too. Stop trying to detect the fake after it lands on your desk. Make the real thing visible while it is being made.
Writing is a process, not a product, and the process leaves a trail. A genuine essay is drafted, stalled on, reordered, second-guessed, and revised. It has a history. A paper generated in one motion does not have that history, and no amount of polish on the final paragraph can fabricate the fifty minutes of thinking that a real draft contains. Process-visible writing, where the drafting itself is the evidence, is the educational version of an honest supply chain. You are not squinting at the tube. You are watching it get filled.
This is also the more humane move, and the more educational one. Forensic detection treats every student as a suspect to be cleared. Process visibility treats writing as something worth seeing happen. It rewards the productive struggle, the false starts, the slow assembly of an idea, which is the part that was always the point. The struggle is not a bug in the assignment. It is the assignment.
What a teacher can actually do on Monday
You do not need a supply chain consultant to bring this thinking into a classroom. A few practical moves:
Treat a clean final draft the way you treat a suspiciously cheap luxury product. Not with an accusation, but with a question: can you show me where this came from? Ask to see the drafting, the outline, the messy middle.
Build assignments that have a paper trail by design. In-progress check-ins, draft milestones, and reflections on what changed and why are not busywork. They are provenance.
Be deeply skeptical of any tool that hands you a probability and asks you to act on it. A detector score is a marketing claim, not a lab result. If you would not let a six-dollar scanner decide whether to trust a person, do not let a paragraph-classifier do it either.
And talk to students about authenticity honestly. They live in the dupe economy. They already understand that something can look perfect and be hollow. The same is true of an essay a machine wrote for them. It can score an A and contain none of their thinking, which means they paid full price and got nothing inside the tube.
AI in schools is not going away, and pretending we can ban it or reliably sniff it out after the fact is the educational equivalent of memorizing logo fonts. The fakes will keep getting better. The only durable answer, in the makeup aisle and in the classroom, is to stop authenticating the surface and start making the source visible.
Mirror, mirror on the wall: the realest one was never the one that looked perfect. It was the one that could show its work.