Substack introduces AI detection feature
Substack, in partnership with Pangram, has launched a tool that lets readers check how much of a post was written by AI.
Substack has announced a new AI detection tool developed in partnership with Pangram. The tool allows readers to scan any Substack content to assess how much of it was generated by artificial intelligence. It works on text longer than 100 words published from today onward.
Additionally, Substack is adding a new space for creators to explicitly disclose whether and how they used AI in their content. The AI detection capabilities are available today on the web and iOS, with Android support coming later.
The blog post announcing the feature takes a surprisingly critical tone, taking a swipe at LinkedIn for the abundance of AI-generated content on that platform and coining the term "Claudefishing" for attempts to fake human connection with AI slop. However, Substack acknowledges that even the best AI detection tools cannot guarantee accurate assessments. The company notes the limitations of Pangram's detection and hints at other features under consideration regarding AI content and preferences. Substack emphasizes that these measures are meant to set expectations for readers, summing up its stance as "people should know what they're getting."
