Substack Adds Pangram-Powered AI Detection, Drawing Creator Backlash
Key Takeaways
- ▸Substack integrated Pangram's AI detection tool as a built-in feature allowing users to scan any piece of writing for AI-generated content
- ▸Creators argue that AI detection tools are fundamentally unreliable and discriminate against legitimate uses of AI for editing and writing assistance
- ▸Pangram claims a false positive rate of roughly 1 in 10,000, but researchers and industry professionals dispute the real-world accuracy of AI detectors
Summary
Substack announced the integration of Pangram's AI detection tool, which scores content on a percentage basis to indicate how much of a piece was AI- versus human-generated. The platform partnered with Pangram to help readers understand the nature of the content they're consuming, with CEO Chris Best stating that automated detection combats AI-generated content that "pollutes the commons." However, the feature has sparked significant backlash from newsletter creators who characterize it as a "witch hunt." Multiple creators and digital writing professionals argue that AI detectors are notoriously inaccurate and that false positives could irreversibly damage their reputations, particularly affecting non-native English speakers and neurodiverse writers.
- Critics warn that false accusations of AI use could tarnish creator reputations, with particular concerns for non-native English speakers and neurodiverse writers who may be flagged incorrectly
- Substack introduced a "How I make this" disclosure feature as an alternative, allowing creators to voluntarily explain their writing process and AI usage
