Substack Introduces AI Detection Feature to Flag AI-Written Content

Substack has launched a new tool that lets users check whether newsletters, posts, comments, and replies on its platform were written by AI. The feature integrates AI-detection software from Pangram and provides an estimate of how much of a given piece of content was human-written versus AI-generated.

Substack CEO Chris Best said the move represents a positive use of AI, explaining in an online exchange with Pangram founder Max Spero that writers should focus on the harder, more valuable work of developing original ideas while software handles supporting tasks. The company said the tool is not intended to penalize AI-assisted writing but to encourage transparency, and it will allow writers to add an optional AI authorship note disclosing their process.

The feature applies to any post, note, reply, or comment exceeding 100 characters within Substack’s app. Publishers will also be able to run the detection tool on their own drafts before publishing and can flag or remove scans they believe are inaccurate.

The launch places Substack among a growing number of platforms adopting AI-labeling practices, following similar moves across social media and music streaming services to identify AI-generated content for users.