Substack has introduced a new feature designed to help users identify content that may have been generated or assisted by artificial intelligence. The tool, developed in partnership with AI detection company Pangram, analyzes posts, notes, replies, and comments to estimate the proportion of text that could be AI-created or produced with AI assistance. The feature is currently available on the web and iOS versions of the platform, with an Android release expected in the near future.
Readers can access the detection tool by selecting the “Scan for AI text” option from the menu at the top of individual posts. The analysis function applies only to content exceeding 100 words in length. According to the company, the integration aims to address growing concerns about AI-generated material on digital platforms by promoting greater clarity about content origins.
Substack co-founder and CEO Chris Best framed the initiative as a response to shifting reader expectations in an era of increasing AI adoption. He emphasized that the primary concern is not AI usage itself, but rather situations where readers unknowingly engage with content created entirely by algorithms without human involvement. The company will also introduce a “How I make this” statement feature, allowing writers to describe their creative process and provide context about their use of tools.
Best acknowledged limitations in the detection technology, noting that Pangram can identify whether AI played a role in text creation but cannot assess the quality of human effort invested or distinguish between AI used as a research tool versus content generation. Writers will have the ability to scan their own drafts and report cases where they believe detection results are inaccurate. The platform frames the effort as essential for maintaining trust in creator attribution and protecting the viability of writing as a profession.