Substack · 2026-07-22 · notable
Substack ships AI-writing detection — Pangram scores every post and comment
Substack partnered with Pangram to score any post, note, reply, or comment as human, AI-assisted, or AI-generated. Readers tap the three-dot menu on any item longer than ~100 characters. Web and iOS today, Android later.
Substack readers can now check whether a post, comment, or Note was written by a human, by AI, or somewhere in between — powered by Pangram.
What is it?
Substack's AI detection feature is a per-item, on-demand scan powered by writing-detection vendor Pangram. It shipped on 22 July 2026 on Substack's web and iOS apps, with Android to follow. Any text-item longer than about 100 characters can be scored.
How does it work?
Readers tap the three-dot menu on a post, Note, reply, or comment and pick the analysis option. Pangram returns a percentage breakdown — human, AI-assisted, or AI-generated — for content published from late July 2026 onward. Substack also added a creator statement space so writers can voluntarily disclose how they used AI.
Why does it matter?
Newsletter platforms have avoided taking a public position on AI-written writing. Substack's move gives readers a first-line signal without banning generative tools, and gives creators a way to compete on provenance. Pangram itself warns it is suggestive, not definitive: precise, clinical human writing can trigger false positives.
Who is it for?
Substack readers curious about the origin of what they subscribe to; creators wanting to signal that their work is human-written
Try it
Open any long Substack post on web or iOS, tap the three-dot menu, and run 'Detect AI'