In 5 years, nobody will give a damn about AI-detectors.
WESTENBERG
SubscribeSign in
In 5 years, nobody will give a damn about AI-detectors.<br>Why the Pangram screenshot is a doomed beast
JA Westenberg<br>Jun 29, 2026
29
Share
Every time an article goes viral, the Pangramguy appears. He tags Pangram and asks for a score, or he pastes away and smugly takes a screenshot of the output, watching the number climb toward “100% AI generated” with the satisfaction of a hall monitor catching a kid with a bag of weed.<br>This is fake. This is lazy. This is entirely beneath consideration.<br>For the last 12 months, that screenshot has carried social weight - it’s an excellent tool to shame anyone for their writing when you either disagree with their work or don’t like them personally. But that social weight is already lifting, and within five years, the performance will look dated and damn near quaint.<br>This whole apparatus is a transitional artifact. Tools like Pangram will get better and better, but the actual verdict will matter less and less. The share of people who give a shit about whether a given piece of writing, art, or code was produced with AI is going to fall off a cliff, and that fall has already started.<br>Five years from now, asking “Is this AI?” will feel about as urgent and meaningful as asking whether a photo was taken on film or digital. A few specialists, myself included, will care intensely. Everyone else will have moved on. I’m not making a values or morality-based call on whether this is a good or bad thing.<br>I’m saying it’s inevitable, either way.<br>People will just stop giving a shit - the way people always stop giving a shit about change and technological advances, which is to say gradually, and then all at once, without ever actually making an active decision.<br>The detectors were always going to go the way of the buffalo.
Start with the why: why does anyone run an AI detector in the first place? Nobody wakes up with a pure // disinterested curiosity about the provenance of a block of text; the detector, and the act of using it, are a proxy for something else they actually care about - usually one of three things:<br>Did this person put in the right amount of effort to produce thing?
Is thing any good?
Can I trust the person who created thing?
For a relatively brief window (relative to asteroids and planet death, etc.), “written by a human” was a decent enough proxy for all three. If a student wrote their own essay, they probably engaged with the material, at least insofar as they searched Amazon Books and glanced at a few previews. (I was a bad student. Sue me.) If a writer produced their own prose, it reflected actual thought, regardless of how much of that thought belonged to the writer himself. If a candidate wrote their own cover letter, it told you something about their effort and competence, no matter how much of it was likely bullshit. Authorship was both easy to check and only weakly correlated with what you actually wanted.<br>The “is this AI” question was never the question; it was a shortcut around better questions. When the correlation breaks and the proxy decouples, people will abandon it. It has simply stopped being a useful measurement, and useful measurements get dropped without ceremony.<br>When enough of the internet is going to be AI-generated (let’s not flatter ourselves with optimism that the reverse will be true), and you can assume you’re reading AI content at least half the time, the test of AI-or-not becomes largely useless. Post-AI, you actually have to measure each piece and each originator, on a case-by-case basis, against a set of rules that define your answer to those three questions, which is probably a better outcome than using a shortcut in the first place.<br>But what about the process? What about the act of sitting down at a typewriter and bleeding?<br>Suppose I hand you a perfect provenance oracle and a stack of writing. You run every piece of content through it, and you now know, with absolute certainty, the origin of each piece. What have you actually learned?<br>You’ve learned a series of facts about a manufacturing process; you haven’t learned anything about quality, truth, effort, or whether any piece is actually worth your time - unless your sole decision-making factor is based on that manufacturing process.<br>None of the millions of people who buy James Patterson novels actually care that James Patterson hasn’t written a James Patterson novel for the best part of a decade; he simply hands outlines and ideas to a ghost writer who produces the finished work. This is a widely known and accepted fact in literary circles, but it just doesn’t seem to matter. Every book that comes out under his name sells like wildfire, regardless of whose fingers touched the keys.<br>For the vast majority of people, and the overwhelming majority of the writing they actually consume, the manufacturing process is not actually a property they care about for its own sake. An airport thriller, a recipe,...