Substack Integrates Pangram to Launch AI Detection

Substack has announced a new AI detection feature for platform articles, replies, and comments through an integration with Pangram. Users can now scan content within the app, and the system will estimate the ratio of human-written to AI-generated text to identify low-quality, inauthentic "AI slop." This product update reflects a direct response from publishing and community platforms to the growing volume of machine-generated text.

Confirmed

Substack has officially integrated detection technology provided by Pangram. Users can scan posts, replies, and comments within the app, and the system will assess whether the content is human-written or AI-assisted. Substack CEO Chris Best emphasized that the core issue the platform aims to address is inauthentic "slop," rather than the use of AI technology itself.

Why it matters

Blogger @peterwildeford views this move as an excellent case study of the d/acc (defensive acceleration) philosophy. Blogger @StewartalsopIII compared Substack's content labeling mechanism to social platform X enabling country/region location tags, suggesting such mechanisms could profoundly alter the platform's existing content ecology and user interactions. Meanwhile, blogger @D3VAUX pointed out that the launch has sparked discussions regarding the accuracy and false positive rates of AI detection tools.

2026-07-22 ~ 2026-07-23 · 10 related posts

Primary sources

5 near-duplicate retellings: sahilypatel · danielrock · peterwildeford · gaganghotra_ · StewartalsopIII