Substack Introduces AI Detector to Enhance Transparency on the Platform

A new tool is being rolled out by Substack, a platform that allows writers and creators to share their work with readers. This tool can scan posts, notes, replies, and comments to estimate how much text may have been written using artificial intelligence or AI assistance.

The detection system uses an external company called Pangram, which specializes in identifying AI-generated content. The integration of Pangram’s technology is aimed at increasing transparency on the platform as AI becomes more prevalent online.

According to Chris Best, Substack’s cofounder and CEO, the primary issue with AI-generated content is not its quality but rather the mismatch between a reader’s expectations and reality when they unknowingly invest their attention in something created without human thought. This phenomenon has been referred to as ‘Claudefishing’.

The new tool can analyze content longer than 100 words by allowing readers to choose the option to scan for AI text from the three-dot menu in the top-right corner of a post.

Substack is also introducing an optional feature that allows creators to explain their writing process with a statement called ‘How I make this’. This addition aims to provide more context and transparency about the content being shared on the platform.

Writers can now use Pangram’s technology to scan their drafts, giving them the option to report any inaccurate results. However, it is essential to note that Pangram can only detect whether AI was used in creating the text, not its quality or the level of human care involved.

Chris Best emphasizes that the goal of this integration is to make it easier for readers to choose where they invest their time and attention. He argues that when readers are unsure about the authenticity of content, it undermines trust in authorship and can have negative consequences for writers who use AI tools responsibly.