TechCrunch: Substack adds AI detection to newsletter posts via Pangram

Substack integrated AI-detection software from Pangram on July 22, 2026, adding a layer of transparency to newsletter content that the platform had not previously offered. The tool scans posts, comments, and replies — anything above 100 characters — and estimates how much of the text was written by a human versus an AI tool.

The feature works in both directions. Readers can run Pangram scans on any newsletter they follow to see the AI-content estimate. Writers can scan their own drafts before publishing and can dispute results they believe are inaccurate. Substack CEO Chris Best framed the addition as consistent with AI handling routine tasks while writers focus on creating ideas, though the practical effect for AI-heavy newsletters is exposure rather than encouragement.

The move follows a pattern playing out across content platforms. YouTube labels AI-generated video, TikTok flags AI content, and music services label AI tracks. Substack’s Pangram integration extends that transparency layer to long-form newsletter writing, a category that has seen significant AI adoption among smaller, individual publishers. For now the detection is informational rather than enforcement — it surfaces estimates, not verdicts, and Substack has not announced penalties for high AI-content scores.

Writers producing newsletters with AI assistance should treat this as a signal about reader expectations rather than a rule change. The presence of a detection tool creates a visible surface where AI reliance can be evaluated, and sustained reader trust is more likely to follow from disclosed, thoughtful AI use than from AI content that merely passes a scan.