You’ve read the essay. You know the one. The structure is perfect. The grammar is flawless. The argument is… nothing. It says nothing. It feels like nothing. It’s the academic equivalent of a meal that fills your stomach but leaves you starving.
That’s what one Reddit user called it: “watered down slop. just like momma used to make.” They were being sarcastic. They were also being right.
Now universities are pinning their hopes on AI watermarks—invisible signatures embedded in AI-generated text that supposedly reveal when a student used ChatGPT. It sounds like a solution. It’s actually a surrender.
Watermarks don’t detect thinking. They detect laziness—and only the lazy kind.
Here’s the problem nobody in administration wants to acknowledge: the watermark will tell you a student used AI. It won’t tell you how. It won’t distinguish between the student who used ChatGPT to brainstorm, refine, and challenge their own ideas, and the student who typed “write me a 2000-word essay on the French Revolution” and hit submit.
Both get flagged. Both get punished. Or neither does—because the same AI that creates watermarks can be asked to remove them. The paradox is almost poetic: universities are trusting the technology they don’t trust to catch the students they can’t evaluate using the methods that no longer work.
But that’s not even the worst part.
The real danger isn’t that watermarks will fail to catch cheaters. It’s that they’ll incentivize a new breed of cheater—one who produces ‘good enough’ slop that passes every check.
Think about it. If the watermark is the gatekeeper, students will optimize for passing the watermark, not for producing original thought. They’ll learn which tools evade detection. They’ll learn how to paraphrase AI output just enough to strip the signature. They’ll become experts in the art of looking like they learned something.
The watermark doesn’t raise the bar. It moves the bar sideways and calls it progress.
We’ve seen this movie before. Standardized testing was supposed to ensure accountability. Instead, it created teach-to-the-test factories where students learned to fill bubbles, not think critically. Plagiarism detectors like Turnitin were supposed to protect academic integrity. Instead, they turned every essay into a game of synonym roulette—swap enough words, beat the algorithm, call it original.
Every technical fix for a human problem creates a new way to game the system. Watermarks will be no different.
You cannot watermark a mind. You cannot detect the absence of thought with a regex.
What universities actually need to do is harder than any watermark. They need to ask: why are we assigning work that AI can complete in the first place? If an essay can be written by a language model in 12 seconds, what exactly was it measuring?
Was it measuring research ability? Then change the assessment. Was it measuring critical thinking? Then make the student defend their work out loud. Was it measuring writing skill? Then watch them write—in the room, with a pen, with no screen between them and the page.
The universities that survive the AI wave won’t be the ones with the best detection tools. They’ll be the ones that rethink what assessment means when knowledge is free, instant, and infinite.
The ones that don’t? They’ll keep assigning the same essays. They’ll keep running the same watermarks. And they’ll keep reading the same watered down slop—perfect grammar, perfect structure, perfectly empty.
Just like momma used to make.
FAQ
Q: Can't watermarks just be improved to be more accurate?
A: Accuracy isn't the problem. The problem is that a watermark tells you AI was used, not whether it was used ethically. A student who uses AI as a thinking partner and a student who uses it as a ghostwriter look identical to any detection system. Better watermarks just create better false positives.
Q: So what should universities actually do right now?
A: Stop assigning work that AI can complete instantly. Move to oral defenses, in-class writing, project-based assessments, and assignments that require personal experience or real-time collaboration. If your assessment can be passed by a language model, your assessment is the problem.
Q: Isn't this just the same panic educators had about calculators and Wikipedia?
A: No. Calculators automated computation; Wikipedia automated reference-finding. AI automates synthesis—the actual thinking part. When the tool that replaces human effort is the same tool that replaces human thought, the comparison breaks down entirely.