Substack launches AI text detector to expose ghostwritten posts

Substack is introducing a new feature that lets readers check whether the content they're viewing was written by humans or generated through AI tools. The detector scans posts, notes, replies, and comments to flag text that may have been created with artificial intelligence or substantial AI assistance.

How the tool works#

The feature, powered by AI detection company Pangram, is now available on the web and iOS app, with Android support coming soon. Readers can activate a "Scan for AI text" option from the three-dot menu on any post containing 100 or more words. The system then estimates what portion of the text could be AI-generated.

Substack also introduced a complementary feature allowing writers to disclose their creation process through a "How I make this" statement. Creators can preview their own drafts using Pangram before publishing and can dispute inaccurate detection results.

Why this matters#

CEO Chris Best framed the initiative as addressing a credibility crisis rather than condemning AI use outright. The real issue, Best argues, arises when readers unknowingly consume entirely AI-generated content while expecting human authorship—what he termed "Claudefishing." He emphasized that "when readers have to wonder if what they're reading is real, it undermines trust in authorship and threatens the livelihood of writers."

The move reflects growing concerns about AI-generated content flooding social media platforms and eroding audience confidence in authentic human voices.

Tool limitations#

Best acknowledged that Pangram has constraints: the technology can detect whether AI was involved in text creation but cannot assess whether human effort or editorial care shaped the final product. It also cannot identify whether AI tools served as research sources rather than primary writers.

The integration signals Substack's commitment to transparency as artificial intelligence becomes increasingly embedded in content creation workflows.