Substack Just Gave Readers a Way to See How Much Is Written by AI — And That’s a Good Thing

I have a confession to make. As someone who writes for a living and runs a blog that publishes multiple times a day, I think about AI writing tools a lot. Not just how to use them effectively, but the bigger question that keeps me up at night: how much AI is too much, and should readers get to know?

Substack just made that question a lot more concrete. This week, the newsletter platform launched an integration with Pangram, an AI writing detection tool, that lets readers scan any post, comment, or reply to see an estimate of how much was written by a human and how much was generated by AI.

How It Works

Available in the Substack app for any content above 100 characters, the feature uses Pangram’s detection engine to give readers a breakdown of human versus AI writing. It applies to posts, notes, replies, and even comments. Substack is also giving writers the option to include an AI author’s note that properly discloses their use of AI tools.

CEO Chris Best framed it as a feature that’s actually good for writers, not a gotcha for readers. “Software should do everything else, but I think you do want the person to do the hard part,” he said in an online discussion with Pangram’s founder. The idea is that AI can help with research, editing, and polishing, but the core idea, the unique perspective, the voice — that should still come from a human.

The Transparency Wave

Substack isn’t the first platform to go down this road, and it definitely won’t be the last. We’ve seen YouTube start automatically labeling AI-generated videos. Spotify and Tidal have updated their AI policies to label and even penalize AI-generated music. The broader push across the internet is clear: platforms are moving toward mandatory or semi-mandatory transparency about AI involvement in content creation.

This is part of a larger reckoning with what AI means for trust and accountability online. If you’ve been following the conversation around AI responsibility, you know this tension has been building for a while. The question of who owns what an AI creates is still being fought in court, as we saw with the xAI lawsuit where the company actually sued its own user over content Grok generated. And researchers have been documenting how over-reliance on AI tools can make people less accurate while paradoxically making them more confident — what I covered as the “cognitive surrender” problem.

Why This Matters for Independent Publishers

Here’s the thing. If you run a blog like I do, or a newsletter, or any kind of independent publication, this tool affects you whether you use AI or not.

If you don’t use AI, this is a huge opportunity. The Substack feature is essentially a trust signal. Readers who see a high “human-written” score on your content will know you put in the work. In a sea of AI-generated content, being transparently human could become a competitive advantage.

If you do use AI — and let’s be honest, most of us in tech use it for something — this tool forces you to think about how you present that relationship. Substack’s approach is actually quite reasonable: they’re not banning or penalizing AI-assisted writing. They’re giving writers the option to add a “how I make this” statement, explaining their process. The transparency is meant to build trust, not destroy it.

The Elephant in the Room: False Positives

Let me address something that everyone in the AI detection space tiptoes around. AI writing detectors are not perfect. They have a well-documented problem with false positives, especially for non-native English speakers, writers with distinctive styles, and technical content that uses formal language.

Pangram is better than most, according to what Substack has shared, but no detection tool is 100% accurate. A writer who uses AI for grammar checking and headline brainstorming might show as partially AI-assisted, even if every sentence is originally theirs. This is why Substack lets publishers request re-scans and remove scans from their work that they believe are mistakes.

The imperfect nature of these tools is exactly why the approach matters more than the technology. Substack’s decision to make it an informational tool rather than a punitive one is the right call. It gives readers context without turning the feature into a weapon.

What This Means for the Blogging Landscape

I see this as a natural next step in the evolution of online content. A few years ago, nobody talked about AI in their writing process. Then it became something people mentioned in passing. Now platforms are building transparency features directly into their products.

This shift toward transparency isn’t happening in isolation. We’re seeing regulators push for AI accountability across every industry, from finance to healthcare to education. The push for an AI equivalent of financial regulators — like what DeepMind’s CEO proposed with a FINRA-style body for AI — is gaining traction. And in Europe, regulators are forcing platforms to open up their ecosystems to rival AI services.

All of this points in the same direction: AI involvement in content creation is becoming something that must be disclosed, not hidden.

Practical Advice for Fellow Publishers

If you run a blog, newsletter, or any content site, here’s what I’d suggest doing right now:

• Be proactive about your AI usage disclosures. Don’t wait for a platform to force it — tell your readers how you work. It builds trust.
• Run your own content through an AI detector before publishing. If there are patterns you don’t like, you can adjust them before readers see them.
• Focus on what AI can’t do: original analysis, personal stories, unique perspectives, and genuine voice. Those are your moat.
• Keep an eye on how other platforms respond. If Substack’s feature is well-received, expect WordPress, Medium, Ghost, and others to follow.
• Audit the AI tools you use. Not all AI writing assistance is created equal. Some tools are genuinely helpful for productivity, while others are essentially content mills in a different package.

Bottom Line

Substack’s AI detection tool isn’t the end of independent publishing. It’s the maturing of a relationship that has been evolving ever since large language models became accessible to the public. Readers deserve to know what they’re reading. Writers deserve to be transparent about their process. And platforms that build bridges between those two goals are doing the right thing.

I, for one, welcome the transparency. It pushes me to be more thoughtful about how I use AI tools in my own workflow, and it reminds me that the real value I offer isn’t in generating text at scale, but in having something worth saying. That’s the hard part. And honestly? It should be.