Skip to main content

Substack rolls out AI text detection, powered by Pangram

Substack, the popular newsletter platform, has officially rolled out an AI content detection feature. Starting July 21, 2026, readers on the web and iOS app can now check whether a piece of writing was crafted by a human or churned out by an algorithm. An Android version is also on the way.

Powered by a tool called Pangram, the feature works across all types of content on Substack — articles, notes, replies, and comments. To use it, readers simply click the “…” menu on any post and select “Scan AI Text.” A pop-up then shows the breakdown: what percentage is human-written, what's AI-generated, and what's a mix of both.

But there's a catch: the tool only works on texts published after 8:30 AM on July 21, and only on paragraphs longer than 100 words. That's because Pangram's underlying technology is a classifier neural network trained on data collected entirely before 2021 — back when generative AI wasn't yet everywhere. This helps keep the detection criteria objective.

Image

While giving readers more insight, Substack hasn't forgotten its creators. Writers can test their own drafts or add creation notes explaining how they wrote a piece. They can also disable the detection feature entirely or remove the AI label at any time. So it's not about policing — it's about choice.

Chris Best, co-founder and CEO of Substack, explained that the move addresses a growing problem: “mismatch between reader expectations and actual content.” With large language models making content production easier than ever, the share of AI-generated text online keeps climbing. This feature is Substack's way of balancing the convenience of AI tools with the value of human-created work.

Key Points

  • Feature launch: AI content detection now live on web and iOS, Android coming soon.
  • How it works: Click “…” menu, select “Scan AI Text” to see human/AI percentages.
  • Limitations: Only applies to texts published after July 21, 2026, and paragraphs over 100 words.
  • Creator control: Writers can test, add notes, or disable detection entirely.
  • Why it matters: Helps readers know what they're reading and supports human originality in an AI-filled landscape.