Substack just gave readers a way to check if their favorite newsletters are written by AI or by humans. The platform rolled out an AI detection feature on July 22 that scans posts, notes, replies, and comments in its mobile app and displays an estimate of how much content was AI-generated. The move is powered by Pangram, an AI detection company founded by Max Spero, and marks the first time a major publishing platform has built AI transparency directly into its reading experience.
But as a founder or newsletter writer, you need to ask the hard questions. How accurate is this thing? Can it tell the difference between AI-assisted editing and AI-generated slop? And how does it compare to standalone tools like GPTZero or Originality.ai that you might already be using?
Here is what founders need to know about Substack's AI detection feature, how it actually works, and whether it should change how you approach content creation on the platform.
Why It Matters
Trust is the currency of independent publishing. Substack has built its entire business model around writers building direct relationships with paying subscribers. If readers cannot tell whether a newsletter they pay for was written by a human or spat out by Claude, that trust breaks.
Substack CEO Chris Best told TechCrunch the feature is about transparency, not punishment. Writers can optionally add an AI author's note explaining their process. The goal is to nudge creators toward disclosure rather than penalize them. But the reality is that once this feature is live, every reader with the Substack app can check your work.
The timing is not accidental. Google's scaled content abuse policy explicitly targets AI-generated content that adds no value. Substack newsletters that get flagged as mostly AI-written could lose search traffic. And paying subscribers who discover their $10/month newsletter is 80% AI-generated will likely cancel.
How Substack's Tool Compares to the Alternatives
Substack's detection is not a standalone product. It is a lightweight integration built into the mobile app. By contrast, GPTZero and Originality.ai are purpose-built detection platforms used by publishers, educators, and content teams. Here is how they stack up:
| Feature | Substack (Pangram) | GPTZero | Originality.ai |
|---|---|---|---|
| Platform | Built into Substack mobile app | Web app, Chrome extension, API | Web app, Chrome extension, API |
| Scan scope | Posts, notes, replies, comments | Any text you paste or URL you submit | Any text, URL, or bulk scan |
| Min length | 100 characters | 250 characters recommended | 50 characters (less accurate below 100) |
| Batch scanning | Not available | Up to 50 files in batch mode | Unlimited batch scanning with team plans |
| Pricing | Free (included with Substack) | Free tier (limited), Pro from $15/month | Free tier (200 credits), paid from $12.95/month |
| API access | Not public | Available (paid) | Available (paid) |
| False positive rate (claimed) | Not disclosed | Less than 2% (self-reported) | Less than 1% at 95% confidence (self-reported) |
The biggest difference is context. Substack's tool only works within its own ecosystem. GPTZero and Originality.ai scan anything you throw at them. Substack also has no API and no batch scanning, which means you cannot run your entire newsletter archive through it before publishing.
What This Means for Founders
If you run a newsletter on Substack, you need to take this seriously. The detection feature is live in the app right now. Any reader can check any post. If your newsletter shows up as 70% AI-generated, that fact is visible to every paying subscriber.
Here are three things to do today.
- Check your own newsletters. Open the Substack app, find one of your recent posts, and see what score it gets. Do this before a reader does it for you. If the score is high, reconsider how much of your writing pipeline relies on AI generation versus AI assistance.
- Add an AI author note. Substack lets writers optionally include a disclosure about their AI use. If you use AI for editing, research, or drafts, say so explicitly. Transparency builds trust. Hiding it erodes it.
- Use a dedicated detector for pre-publish checks. Substack's tool is reactive - readers see the score after you publish. Use Originality.ai or GPTZero before hitting publish to catch surprises. The free tiers are enough to spot-check your most important posts.
The harder question is what constitutes acceptable AI use. Using AI to fix grammar or suggest better phrasing is very different from pasting a prompt and publishing the output verbatim. Substack's tool is not sophisticated enough to make that distinction, nor should it. But readers will form their own judgments regardless.
Bottom line: Substack's AI detection is a meaningful step toward content transparency on the platform, but it is not a replacement for dedicated detection tools. If you publish frequently and care about maintaining reader trust, use Substack's in-app scan as a baseline and run pre-publish checks through GPTZero or Originality.ai for anything important. And if your newsletter gets flagged at above 50% AI content, take that as a signal that your readers might be better served by more human effort and less automation.
Disclosure
Some of the links in this article are affiliate links. If you sign up through these links, we may earn a small commission at no extra cost to you. This helps us keep The Break Daily free for everyone.