Substack has begun rolling out AI transparency features built in partnership with Pangram, the AI-detection company, giving readers a way to see how much of a given post, Note, reply, or comment was likely written by a human versus produced with AI assistance. The tools are live now on web and iOS, with an Android rollout to follow.

The feature works on text longer than 100 words and surfaces an estimate of AI involvement only when a user requests a scan. Substack is also introducing an optional “How I make this” statement that lets creators disclose their process and set expectations for readers up front. Creators can run Pangram on drafts before publishing and can flag or remove scan results they believe are inaccurate.

Source: “Against Claudefishing” – Chris Best, July 21, 2026

Substack framed the launch as a response to what co-founder and CEO Chris Best called “Claudefishing,” his term for the gap between a reader’s assumption that a human wrote something and the reality that no one did. “The problem is when there is a mismatch between a reader’s expectation and reality, especially when they unwittingly invest their attention in something with no human thought on the other end,” Best wrote in a post announcing the features.

Best said the company is not trying to discourage AI use outright. “We’re not against people using AI to assist their work, and we think people should be free to choose which tools they use to express themselves,” he wrote, adding that Substack uses AI internally for software development, research, and product features like clipping and translation. He said the company is considering further tools, including community-level AI preferences set through Reply Rules, more ways for creators to signal the value of their work, and reader-side controls over what gets recommended to them.

Pangram’s own data helped shape the timing. A report the company published this month, drawn from more than a million posts scanned through its Chrome extension since April, found that AI-generated content is concentrated in longform writing across social platforms, with a quarter of posts over 250 words flagging as fully AI-generated. 

Source: Pangram

LinkedIn had the highest AI share of any platform studied, with more than 40% of its longform posts flagging as fully AI-generated, followed by X, where nearly half of longform articles showed some degree of AI involvement. Substack had the lowest combined AI rate among the longform platforms Pangram measured, though more than a fifth of its posts still flagged as AI-generated or AI-assisted.

Pangram said its detection model, Pangram 3.3, achieves a 0.01% false positive rate, though Best noted the tool can only estimate whether AI was used, not whether a post reflects genuine human care or was sourced from AI-generated material.

Subscribe to Our Newsletter

Check Out Our Podcast