You write a newsletter. You spend hours on it. You pour yourself into every paragraph. And now, a machine is going to tell your readers whether you’re “real” enough.
Welcome to Substack’s latest move: an AI detection tool that scans newsletters and flags content it suspects was generated by artificial intelligence. The platform that built its brand on writer independence and editorial freedom is now running a surveillance system on the very people who made it relevant.
The irony isn’t subtle — it’s the point.
Here’s the setup: Substack uses AI to detect AI. The platform that has actively promoted AI-assisted writing features, that has integrated AI tools into its own ecosystem, is now positioning itself as the arbiter of authenticity. It’s like a casino installing breathalyzers at the exit doors.
But the irony is the shallow critique. Everyone sees it. The comment sections are already full of people pointing out the obvious — “so they use AI slop to accuse others of using AI slop?” — and yes, that’s funny. But it’s also a distraction from what’s actually happening.
The real move here is a power grab disguised as quality control.
Think about it from the writer’s perspective. You publish a piece. Substack’s algorithm runs silently in the background and slaps a label on it — “likely AI-generated” or “human-written.” You have no idea what thresholds triggered that label. You can’t see the scoring. You can’t challenge it meaningfully. But your readers see it, and suddenly you’re defending yourself against a verdict handed down by a black box.
When a platform makes you prove you’re human, it has already decided you’re guilty until cleared.
This is the fundamental shift nobody’s talking about. Substack isn’t solving the AI content problem. It’s creating a new trust infrastructure where the platform — not the reader, not the writer — holds the authority to define what counts as “authentic.”
And here’s where it gets truly insidious: AI detection tools are notoriously unreliable. They produce false positives at rates that should terrify anyone who writes. Non-native English speakers, writers with clean prose styles, people who edit carefully — all of them are statistically more likely to be flagged as “AI-generated” by these systems. The tools don’t detect AI. They detect patterns that resemble AI’s patterns, which often overlap with patterns of competent, structured writing.
So who gets hurt? The careful writers. The polished writers. The writers who happen to sound too clean for an algorithm’s comfort.
Meanwhile, the people actually churning out AI-generated slop will adapt. They already are. Every detection tool that gets publicized becomes a training signal for the next generation of AI writing tools to avoid detection. It’s an arms race, and the detection side always loses, because the generation side moves faster.
You cannot detect your way out of a problem created by the same technology.
What Substack is really doing is more strategic than it looks. By building this tool, they’re creating a new dependency: writers now need Substack’s “clean” rating to maintain credibility. Readers learn to trust Substack’s label, not their own judgment. The platform becomes the gatekeeper of a standard it invented, controls, and can enforce selectively.
Imagine the future scenarios. A prominent writer publishes a controversial piece critical of Substack’s policies. Suddenly, their AI detection score starts fluctuating. A new competitor newsletter platform emerges — and Substack’s tool conveniently flags content referencing it as “suspicious.” This isn’t conspiracy thinking. It’s the logical endpoint of giving a platform algorithmic authority over authenticity.
The burden of proof has shifted entirely. Writers used to be trusted until proven otherwise. Now they’re suspect until an algorithm says otherwise. And the algorithm is opaque, biased, and wrong often enough to make its verdicts dangerous.
Authenticity was never a technical problem. It was an editorial one. And Substack just turned it into a surveillance infrastructure.
If you write on Substack, this should alarm you — not because AI content is a non-issue, but because the solution being offered gives a corporation unilateral power to label your work. If you read on Substack, this should make you skeptical — because the label you’re being shown as “truth” is itself generated by the same class of technology it claims to police.
The real question isn’t whether AI-generated newsletters are a problem. They are. The real question is whether a platform using AI to judge AI, with no transparency, no appeals process, and a clear conflict of interest, is the entity you want defining truth for you.
The answer should be obvious. But Substack is betting it won’t be.
FAQ
Q: Isn't some kind of AI detection better than nothing?
A: No — a broken detection system is worse than no system at all. False positives damage real writers' reputations, create a false sense of security for readers, and give the platform power it shouldn't have. A bad solution often prevents the development of a good one.
Q: What does this mean for me as a Substack writer?
A: Your work can now be silently flagged by an algorithm you can't see, challenge, or appeal. If your writing is clean, structured, or edited well, you're statistically more likely to be falsely accused. Your credibility now partially depends on a black box's verdict.
Q: Isn't this just Substack protecting quality on its platform?
A: If it were about quality, the tool would be transparent, appealable, and open about its methodology. Instead, it's opaque and unilateral. Substack is positioning itself as the sole arbiter of authenticity — a standard it invented, controls, and can selectively enforce. That's not quality control. That's gatekeeping.