Substack has launched a feature that lets users check how much of a newsletter post, comment or reply was written by AI rather than a human, in partnership with AI detection software Pangram. The move could initially unsettle Substack's ecosystem, as it may reveal that many newsletters aren't entirely human-written, potentially denting trust in and the reputation of a platform built around independent, high-quality writing. However, the company frames it as a long-term safeguard against "AI slop," aiming to help readers understand when content is machine-generated and to preserve confidence in what they read.

The detection tool works on any post, note, reply or comment over 100 characters within Substack's app, providing an estimate of the human-versus-AI split. Writers can also add an optional "AI Author's note" to disclose their process, and publishers can run scans on their own drafts before publishing, or flag and remove scans they believe are inaccurate. Substack CEO Chris Best described this as "good use of AI", arguing the technology should handle everything except the essential, valuable work of having an idea worth sharing. The feature follows similar labelling moves by social media platforms for AI images and video, and by music streaming services for AI-generated tracks.