Substack is rolling out a new feature intended to help readers and writers assess whether text on the platform may be AI-generated. Multiple outlets report that the tool scans items such as posts, notes, replies, and comments, and provides an estimate of the portion of the content that could be produced by AI. The company says the feature is described in a blog post and is designed to give audiences additional context when evaluating what they read.
Writers also receive related functionality. Sources say creators can use a pre-publish scan to check their own content before it goes live. Substack also includes disclosure-related options, including a “How I make this” element, which is intended to help writers explain how they created a piece.
One report notes that the scanner is powered by “Pangram,” linking the detector to a specific underlying technology. Across the coverage, the focus remains on providing detection estimates and creator tools rather than changing Substack’s core posting or discovery systems.
