The Irony of "Scan for AI Text"
Why you should be worried about an AI text detector even if you “never” use AI
A lot of people are expressing mixed feelings about Substack’s new AI detection tool. I don’t have mixed feelings. It’s bullshit.
Let me say right out the gate: I am obsessed with AI and use it everyday, but (at this time), I do not use it to write my articles. (I use it for a lot of related tasks in my “content engine” and I’m happy to share what those are in another article).
But I wanted to front load my positioning here. Rather than trying to figure out where to fit AI into my life, I think about the flip-side. What do I not want to use AI for? My actual voice and my presence. In other words, my unique POV and my relationships. My creativity and my connection. These are my “Human Pillars”.
Other than my Human Pillars, everything else is fair game. (And right now I have Claude and ChatGPT working together to create my entire OS. It is exciting.)
Like I said, I don’t use it to write my articles but I made a point to say “at this time”. Because I don’t see a problem with AI writing articles - if AI can actually get the hang of your voice. I’ve certainly experimented myself, but found the results wanting. But a lot of creators say they’ve nailed it - “you just have to train it.” I’m not convinced, but I’m not not convinced.
So I’ve been “quietly” training AI to learn my voice in the background of everything I do for months. From what I can tell it has amassed a whole slew of “don’ts” (e.g., do not use the word “quietly” under any circumstances) and it knows I like an occasional f bomb (but it kind of throws it out at the wrong time, like a kid who doesn’t quite know the appropriate use of a swear word yet). But even beyond these stylistic conventions, it doesn’t know how to hit the way I like to hit. It can analyze my writing but it doesn’t know how to do it the way I do it. So we’re not there yet. But I’m willing to be convinced and I don’t see the problem if it is indeed your voice and your thoughts.
So I’m not against the idea.
What I am against is slop. And guess what? People can write slop too. I’ve read plenty of shitty articles long before ChatGPT came online.
Which is exactly my point - we don’t need a scanner to tell us if it’s good writing or not. You can tell for yourself if an article is garbage, regardless of whether it was written by man or machine.
Instead, it’s going to be used for virtue signaling and witch trials.
Virtue signaling is a problem because, I’m sorry - what’s the virtue here? Again, if you figured out how to get AI to say exactly what you wanted it to say, IDGAF who typed it out. Not using AI is not a virtue. The virtue is expressing the unique thoughts that are in your brain. Which doesn’t preclude assistance.
And the witch trials will be sinister.
It’s like a polygraph - can’t be used in court because it’s junk science but they’re still used by law enforcement to form opinions about people. (Side bar: never agree to a polygraph, no matter what they say)*. We know these things are inaccurate. There will be so many people that get accused of using AI that didn’t. It has a bias in particular against people of color and non-native speakers.
It’s bad enough that this is going to ruin the reputation and maybe careers of creators that get caught in the cross-fires. There’s also the chilling effect this may have on people who decide writing isn’t worth it given the judgmental landscape.
And then consider the creators with neurodivergencies and disabilities who are able to produce work and express their voice thanks to AI. Think about someone who has never been able to sit down and type something but now they can voice to text brain dump everything and get an article out of it that they can share with the world. Even if you tell it to use all of your words, it still has to string sentences together to form something cohesive. It’s going to be their own voice but it will have “quietly” given it enough of an AI signature to get flagged..
But the most chilling thing about this development for me is the idea of relying on AI to tell you if a piece of writing is worth reading.
This is exactly what AI cannot do.
You cannot outsource your judgment, your perspective, your critical thinking skills, your intuition, about a piece of writing. You cannot outsource the way a piece of writing moves you.
And you should not outsource your decision-making power to anyone. And that’s what this scan feature does.
Hyperbolic? Maybe. Call me Cassandra, I don’t care. This is bad, very bad. This is a bad use of AI. (My only hope is that it’s alarming enough that it immediately fails, and this article just looks silly in retrospect.)
Put another way: when a course of action is absolutely dripping with irony, that’s a good indicator to slow down and evaluate whether this actually makes sense.
A little about me….
I’m Niki, a holistic health and Human Design coach. Recovering lawyer turned SAHM turned health obsessed turned questioning midlifer.
I spent years being told my labs were “fine” while my body fell apart, until I connected the dots between stress, the nervous system, and physical symptoms. I realized the stress about the cookie was worse than the cookie.
Human Design gave me the missing piece: a way to understand my own wiring instead of fighting it.
Lately I’m also writing about AI a lot. It’s turned into a real nervous system tool for reducing mental load, and I think women should be fluent in it before it gets built without us.
Looking for more support?
Join my fall waitlist if you’re interested in my 6 month 1:1 container
My paid tier includes a personalized Human Design snapshot report with nervous system insights
Throughline members also get an annual Human Design session with me
*****
Disclaimer: The content in this article is for informational and educational purposes only and is not intended as medical advice, diagnosis, or treatment. Always consult a qualified healthcare professional regarding any medical concerns or before making changes to your health, diet, medications, or treatment plan.
*Not legal advice
"Not using AI is not a virtue. The virtue is expressing the unique thoughts that are in your brain. Which doesn't preclude assistance." That's the whole argument, and I think it's right.
Here's what follows, and it makes the tool worse than unreliable. The scanner can't distinguish between someone who supplied every idea and had a machine phrase them, and someone who took every idea from a machine and retyped it in their own words. The second comes back human. By your own criterion — unique thoughts in your brain — the second is the actual failure. It scores them backwards.
So it isn't measuring the thing badly. It isn't measuring it at all.
Which is maybe why the irony runs deeper than AI-detecting-AI. The scanner assumes provenance settles worth. You've already rejected that premise, which is why you're one of the few people in this argument it can't catch out.
For what it's worth, I write with assistance and say so publicly. You've made that unusually easy to say here.
Hear hear. I’ve also failed to hone my AI to genuinely write “for” me. That said, it massively accelerates the iteration, editing & polishing.
Is the result “AI written?”
I frankly couldn’t care less.