Substack Adds Pangram-Powered AI Detector

Substack has built its publishing business around direct relationships between independent writers and paying readers, making confidence in authorship central to the platform’s value. The spread of generative AI has complicated that relationship as newsletters and social posts can be produced or heavily revised by machines without disclosure, a practice Substack Chief Executive Chris Best has described as “Claudefishing.” The company is treating transparency as a trust issue rather than banning AI use outright.

Substack launched its Pangram-powered “Scan for AI text” feature on July 21, 2026. It analyzes eligible posts, Notes, comments and replies published from that date and estimates the share of text that is human-written or AI-assisted. Scans require more than 100 words. The tool is available through Substack Reader on the web and its iOS app, with Android support due soon, while creators can add statements explaining how they use AI.