Substack has added an AI detection feature that lets readers estimate how much of a post, Note, reply or comment was written by a person or with AI assistance. The optional tool is powered by Pangram and applies to eligible content published from July 21.
The feature does not automatically label every piece of content. Readers must request a scan, and the resulting analysis is visible only to the person who initiated it.
Readers Can Request AI Scans
Users can open the menu attached to an eligible post or comment and select “Scan for AI text.” Pangram then provides percentage estimates for human-written and AI-assisted portions.
Substack’s official support page says the feature works on text longer than 100 words, rather than 100 characters. It is available for posts, Notes, replies and comments published on or after July 21, 2026.
The integration is available through Substack’s web and iOS services, with Android support expected later. Substack has not presented the result as definitive proof of authorship, describing it instead as an estimate.
Writers Can Explain Their Process
Substack is also introducing an optional “How I make this” statement. Writers can use it to explain whether they rely on AI for research, editing, transcription or drafting.
Publishers can run Pangram on their own drafts before publication. They can also report scans they believe are inaccurate and request the removal of results attached to their own work.
Chief Executive Chris Best said Substack is not seeking to prohibit AI-assisted writing. In the company’s official announcement, he argued that readers should understand when a writer has delegated the main creative work to software.
Detection Tool Has Technical Limits
Pangram says its classifier analyses writing patterns associated with models including ChatGPT, Claude and Gemini. The company claims an accuracy rate above 99.98%, though AI detection systems can still produce incorrect results or be affected by heavily edited and paraphrased text.
The partnership gives Substack readers more information when deciding which newsletters and writers to support. It also allows the platform to address undisclosed AI-generated material without automatically penalising writers who use AI during part of their process.
Featured image credits: AltCMO
For more stories like it, click the +Follow button at the top of this page to follow us.