Overview
- Substack has integrated Pangram’s classifier so readers can choose “Scan for AI text” from a post’s menu to see what share of a passage reads as AI-generated, AI-assisted, or human.
- The scanner is live on the web and iOS with Android coming soon and only analyzes passages longer than 100 words and posts made after Substack’s July 21 eligibility cutoff.
- Creators can add a “How I make this” statement, run Pangram on drafts, report inaccurate results, and — according to reporting — disable detection on their posts.
- Pangram says its detector is trained on human-written text gathered before 2021 but both companies warn the tool can return false positives or negatives and cannot tell whether a human provided the core thinking behind a piece.
- The opt-in design hands readers investigative control but also raises risks for authors, since careful or non-native writing can resemble machine output and platform choices could affect trust, discovery, and recommendations.