Substack Integrates Pangram AI Detection to Combat AI-Generated Content
Key Takeaways
- ▸Substack's AI detection tool estimates how much text in a post could be AI-generated or written with AI assistance
- ▸Writers can explain their creative process with a new 'How I make this' statement and verify results with Pangram before publishing
- ▸Pangram's detection identifies AI usage but cannot assess content quality or whether meaningful human effort went into creation
Source:
Summary
Substack has launched an AI detection feature powered by Pangram, enabling readers to identify content that may be AI-generated or AI-assisted. The new tool is available on web and iOS, with Android coming soon, and scans posts, comments, notes, and replies longer than 100 words for AI-generated text by selecting 'Scan for AI text' from the post menu. The integration aims to combat 'Claudefishing' — when AI-generated content is passed off as human-written without reader awareness — and increase platform transparency. Substack CEO Chris Best emphasized that the goal is to preserve trust in authorship and protect writers' livelihoods by enabling readers to make informed choices about their attention.
- The feature addresses 'Claudefishing' — the mismatch between reader expectations and content reality that undermines trust in authorship
