Substack Bestsellers show majority of newsletters free from AI-generated content

3 articles · Updated · User Mag | Taylor Lorenz · Apr 27

An analysis using Pangram's AI-detection tool found that 384 out of 575 top Substack Bestsellers had no AI-generated content in their recent posts.
Technology, Philosophy, and Health categories had the highest AI usage, with 28%, 23%, and 22% of top posts partially or fully AI-generated, while some newsletters publish exclusively AI-written content.
Despite growing AI prevalence online, most Substack readers still prefer human-written content, though a handful of high-volume AI-driven newsletters significantly influence category trends and profitability.

How can human writers possibly compete against automated AI newsletters that are designed to maximize both output and profit?

If top business publications use more AI, what is the long-term risk to brand trust when customers discover this?

As AI increasingly trains on its own writing, are we accelerating toward a 'knowledge collapse' where original human thought vanishes?

Is the real enemy AI content, or are online platforms that reward low-effort 'slop' over human creativity the bigger problem?

An MIT study shows AI use reduces brain activity. What is the long-term cognitive cost of outsourcing our thinking to machines?

Does the new White House AI policy truly protect creators, or does it favor tech giants using copyrighted data for training?