Substack has launched an AI detection features that let you check whether a post, reply, or comment was likely written by a person or generated with AI. The AI content checker is powered by Pangram and is now available on the web and iOS.
You can use it to scan text that is longer than 100 words and published from July 21, 2026 onward. The AI analysis is only shown if you choose to request it.
You can also:
- Scan drafts before publishing to see how Pangram evaluates your writing.
- Add a “How I make this” statement to explain how you created your content and whether you used AI.
- Report or remove AI detection results if you believe they are incorrect.
Pangram estimates whether AI helped generate the text. It does not judge the quality of the writing or determine how much effort went into creating it.
For example, a well-researched article that used AI to improve grammar may still be flagged as AI-assisted. It can also not determine whether AI was only used for research or brainstorming.
Substack says the detector is not perfect, but it is accurate enough to help you make your own judgment about the content you read. It adds that the goal is not to fight AI use but to give readers more context about how content was created.
