Substack has introduced an innovative AI detection feature aimed at identifying content that may have been generated by artificial intelligence. This new tool, powered by Pangram, can scan posts, comments, replies and notes on the platform, providing users with an estimate of how much text might be AI-generated or assisted.

The co-founder and CEO of Substack, Chris Best, emphasises the importance of transparency in the age of AI. He highlights that the primary issue is not the use of AI but when readers' expectations are not met due to a lack of human input, a phenomenon he terms 'Claudefishing.'

Substack users can now access this feature through a simple menu option, allowing them to assess content longer than 100 words. Additionally, creators have the opportunity to explain their writing process via a new ‘How I make this’ statement and can use the Pangram tool to check their drafts.

While the AI detection software cannot determine the quality of human care or the intent behind using AI tools, it aims to empower readers in making informed choices. Best argues that platforms rewarding fake content could lead to a 'race to the bottom,' undermining trust and potentially harming writers who use AI thoughtfully.