It’s Tuesday. A blog post drops. You read it. It feels a little too polished. A little too empty. You’re not crazy.
Substack is finally admitting the elephant in the room. They’re launching a new tool to flag AI-generated content. This isn’t just a checkbox. It’s a scanner. It runs against posts, notes, replies, and even comments. The goal? To estimate how much text was actually written by a machine.
The engine behind this? Pangram. An AI detection company. They are rolling this out on the web and iOS right now. Android users? You’re getting a launch “soon.”
Here is the mechanic: If a post is over 100 words, you can click the three-dot menu. Top-right corner. Look for “Scan for AI text.” It gives you a percentage. An estimate. Not a verdict. Just a probability.
The core problem is not people using AI. Or the quality of its output. The problem is when there’s a mismatch between reader expectation and reality.
Chris Best, Substack’s co-founder and CEO, calls this claudefishing. It’s a sharp term for a fuzzy problem. You invest your attention. You give your time. You expect a human on the other end of the keyboard. You get code. That mismatch? It’s deceptive.
Why Trust Matters More Than Accuracy
Best is clear about what Pangram can’t do. It doesn’t judge care. It doesn’t see if a human refined the draft. It doesn’t know if AI was just used as a source or if the whole piece is synthetic. It detects text generation. Nothing more.
So why bother?
Trust.
When readers have to wonder, “Is this real?” the whole platform rots. It threatens writers. Even those using AI thoughtfully. Even those who believe in their work. If the platform rewards fakeness, we get a race to the bottom. Everyone loses.
How Creators Can Opt In
It’s not just for readers. Creators can also use the scanner. They can test their own drafts before hitting publish. They can see what Pangram sees.
But Substack isn’t just hiding behind a black box. They want transparency. That’s where the “How I make this” statement comes in.
Writers can add a note to their profile or post. Explain the process. Did you write it yourself? Did you use AI? Did you edit it heavily? Users can flag inaccurate scans, too. It’s an imperfect system, but it’s a start.
The logic is simple. When you know how a post was made, you can choose to invest your time. Or you can scroll past. That choice matters.
Plenty of writers use tools to help. Grammar checks. Brainstorming. Structure. That’s not cheating. But hiding it? That’s the issue.
We are in a weird moment. AI is everywhere. It’s in our emails. Our ads. Our social feeds. Substack is trying to carve out a space that feels authentic. Real. Human.
Does a percentage guarantee truth? No. But it adds a layer. It forces a pause. A moment of skepticism. Which might be better than blind trust.
The rollout is happening now. iOS users can try it. Android is coming. Web is live.
How will you react when you see the scan? Will you care? Will you assume the worst?
Or will you just keep reading?
