Substack is introducing a detection tool powered by Pangram to estimate the amount of AI-generated text in posts, notes, and comments. According to CEO Chris Best, the integration aims to increase transparency and combat “Claudefishing,” a practice where creators present AI-generated content as human-authored to deceive readers.
Combating the Rise of “Claudefishing”
The term “Claudefishing” describes a specific deception: when readers invest their attention in content they believe is human-written, only to discover it lacks any human thought. Chris Best, co-founder and CEO of Substack, argues this mismatch undermines the fundamental trust required for a creator economy to function. According to a blog post published by the company, the platform’s new tool scans text to provide an estimate of AI involvement.
Did you know?
Substack’s new feature allows creators to preemptively disclose their process via a “How I make this” statement, providing a verified channel for writers to explain how they utilize technology in their workflow.
Transparency vs. AI Utility
Substack’s approach acknowledges that AI is not inherently problematic. Best notes that many writers use AI tools thoughtfully to enhance their output. The objective of the Pangram integration is not to ban automation, but to ensure readers know what they are consuming. By scanning drafts, creators can gain insight into how their work might be perceived by detection tools and can report inaccurate results if the software flags human-written content as AI-generated.
The platform warns that its tool identifies the presence of AI-generated patterns, not the level of human care or effort behind the writing. This distinction is crucial for writers who use AI as a research tool or a structural aid rather than a replacement for their own voice.
Why Authorship Trust Matters
The shift toward AI transparency is a response to the broader flood of automated content on social media. Best warns that platforms which prioritize volume over authenticity risk a “race to the bottom.” When readers can no longer distinguish between human insight and machine-generated text, the value of independent authorship declines. Substack’s strategy is designed to protect the livelihood of creators by rewarding the human connection that serves as the platform’s primary value proposition.
Frequently Asked Questions
- How does the Pangram tool work? It scans posts and comments to estimate the probability that the text was generated or significantly assisted by AI.
- Can writers dispute the results? Yes, Substack has included an option for creators to report inaccurate results if they believe their work was misidentified.
- Is AI banned on Substack? No. The platform encourages transparency through the “How I make this” statement rather than prohibiting the use of AI tools.
- Does this detect human effort? No. As stated by Chris Best, the tool only detects the use of AI, not the quality or the amount of human care invested in the writing.
Join the conversation: How do you feel about AI disclosure in the newsletters you subscribe to? Let us know in the comments below, or subscribe to our weekly newsletter for more updates on the future of digital media.