Substack adds AI detector to scan posts for machine-generated text

Substack announced Tuesday that it is rolling out a new AI detection tool across its platform. The feature, powered by AI detection company Pangram, can scan posts, notes, replies, and comments to provide an estimate of how much text could be AI-generated or written with AI assistance. Readers access the tool by selecting "Scan for AI text" from the three-dot menu in the top-right corner of a post; the scan works on content longer than 100 words. The tool is available immediately on the web and the iOS app, with an Android release coming "soon."

Substack co-founder and CEO Chris Best explained the rationale in a blog post. He argued that the core issue is not AI use itself or even the quality of AI output, but rather "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 termed this problem "Claudefishing." He emphasized that platforms rewarding undisclosed AI content create a "race to the bottom" that can undermine trust in authorship and threaten the livelihood of writers, even those using AI tools thoughtfully.

Alongside the detector, Substack is introducing a new "How I make this" statement that allows creators to explain their writing process directly to readers. Writers can also use Pangram to scan their own drafts before publishing and have the option to report inaccurate detection results. Best acknowledged a key limitation: "Pangram can only detect whether AI was used to make the text, not whether great human care went into creating it, nor whether AI tools were used as a source." The goal, he wrote, is to make it easier for readers to decide whether to invest their time in a blog post by providing visibility into the role of AI in its creation.

The move reflects Substack's effort to preserve trust on a platform increasingly populated by AI-generated content. Chris Best frames the problem not as AI use itself, but as misalignment between reader expectation and reality — specifically when readers unknowingly consume content with "no human thought on the other end." Best calls this "Claudefishing," signaling concern that without intervention, platforms could spiral into a "race to the bottom" where fakeness is rewarded and human writers' livelihoods suffer. The integration with Pangram, a specialized AI detection company, suggests Substack is treating this as a core feature rather than a minor experiment. By pairing detection with a creator-facing "How I make this" statement, Substack is trying to create space for transparent AI use — writers who use AI tools thoughtfully can explain their process — while helping readers distinguish intentional collaboration from undisclosed automation.