Substack is introducing a new tool to help readers determine whether content was written by a human or generated by AI. The new feature, powered by AI detection tool Pangram, is available on Substack posts, notes, replies, and comments.
The AI detection tool is available now on the web and in the iOS app, with Android support to follow. Users can access the feature by selecting “Scan for AI text” from the “…” menu. The tool can scan text longer than 100 words that was published after 8:30 a.m. PT / 11:30 a.m. on July 21, 2026.
Substack co-founder and CEO Chris Best stated, “Not everything made with AI is slop, and not all slop is made with AI.” He emphasized the importance of aligning readers’ expectations with the content they engage with.
The AI detection tool has specific limitations. It will only work on Substack text passages longer than 100 words, and content published before July 21, 2026, is ineligible for AI detection. Users attempting to scan shorter passages will receive a message indicating insufficient text for accurate detection.
Substack allows authors to add a statement explaining how they wrote their content, which will be visible to readers using the AI detection tool. Writers can also run the detection on their drafts and submit a report if they believe it has classified their content incorrectly. Additionally, authors can disable AI detection on their content regardless of Pangram’s classification.
Pangram employs a classifier neural network model trained to distinguish between human and AI-generated text. Substack is considering adding more AI-focused features in the future based on user interest and feedback.
The announcement included a critique of LinkedIn regarding AI-generated content, using the term “Claudefishing” to describe feigned human connection with AI-generated material.
