Substack has launched a new AI detection tool developed in partnership with Pangram. This tool allows readers to assess how much content was written by AI, applicable to text longer than 100 words. The feature became available today on the web and iOS platforms, with Android support expected in the future.
Additionally, Substack is providing a new statement space for creators to explicitly share how AI was utilized in their content. The company aims to set clear expectations for readers regarding AI-generated material.
The blog post announcing the feature includes criticism of LinkedIn, calling out the prevalence of AI-generated content on that platform. It describes efforts to create a false sense of human connection with AI-generated content as “Claudefishing.” Substack noted that tools for detecting AI-generated content, including those from Pangram, do not guarantee perfect accuracy.
The Atlantic has explored the limitations of AI detection tools and highlighted the inaccuracy of Pangram’s capabilities. Substack acknowledged these limitations and hinted at potential additional features regarding AI content and user preferences in the future. The company stated that its goal is to ensure readers understand the nature of the content they are engaging with, summarizing its position as “people should know what they’re getting.”
