IA · 23 July 2026 · 4 min read
Substack Integrates Pangram to Detect AI Text and Combat 'Claudefishing'
In brief: Newsletter platform Substack has integrated Pangram's AI detection software to help readers identify AI-generated text across posts, notes, and comments. The initiative aims to curb 'Claudefishing'—the practice of passing off raw LLM output as genuine human thought. Alongside reader scans, Substack is equipping creators with workflow disclosure tools to maintain reader trust.
by Team Mocchi's
Substack's New Lens: Scanning Newsletters for AI Text
The rapid proliferation of AI-generated prose across digital publishing platforms has reached a defining moment. Substack, the popular platform for independent writers and newsletter creators, has announced a native integration with Pangram, a specialized AI text detection engine. The feature enables readers to scan posts, notes, replies, and comments longer than 100 words directly on the web and iOS app, with Android support scheduled to follow shortly.
By selecting the "Scan for AI text" option from a post's context menu, users receive an estimated breakdown of how much content was generated or assisted by artificial intelligence. The move positions Substack as an early mover among text-first platforms in establishing granular transparency tools, mirroring similar labeling initiatives for synthetic media across major social networks.
Fighting 'Claudefishing' and Preserving Reader Trust
At the heart of the initiative is a drive to preserve the fundamental trust between creators and their audience. As reported by The Verge, Substack co-founder and CEO Chris Best coined the term "Claudefishing" to describe the deceptive practice of presenting pure LLM output as genuine human thought, tricking readers into spending time on content created without human intent.
Best warned that platforms rewarding low-effort synthetic content risk sparking a destructive "race to the bottom." An analysis by TechCrunch notes that while the feature presents a short-term commercial risk for Substack by exposing AI-dependent newsletters on its platform, the long-term goal is to protect high-quality writers and guarantee transparency for paying subscribers.
Detection Limits and Responsible Author Disclosure
Rather than acting solely as a policing tool, the Pangram integration gives creators proactive control. Writers can run draft scans before publishing, report potential false positives, and publish a dedicated statement explaining their creative process and tool usage.
Substack leadership acknowledges that AI detection software has inherent limitations: Pangram measures linguistic patterns rather than human care, research quality, or editorial oversight. Consequently, the feature is not designed to ban AI tools altogether, but to ensure that readers are never misled about what they are consuming.
Mocchi's take
For businesses and creators navigating corporate communications and content marketing, Substack's move underscores a critical shift. As AI-generated content becomes cheap and ubiquitous, authentic human perspective and transparent workflows become a company's strongest brand assets. We advise organizations to establish clear AI disclosure policies early on and emphasize human expertise in their publishing strategy—because customer engagement and brand authority rely on trust, not uncredited automation.