Two-thirds of top Substack newsletters contain no AI-generated content
A analysis of thousands of posts from Substack's bestseller lists found that the majority of top newsletters remain human-written, but AI-generated content concentrates heavily in specific categories and among a small group of high-volume publishers.
The study examined the 10 most recent posts from the top 25 Substack Bestsellers across all categories using Pangram, an AI detection tool. Of 575 newsletters analyzed, 384 contained no detectable AI-generated writing.
Technology leads in AI adoption among top newsletters. Twenty-eight percent of writing in bestselling Technology newsletters is fully or partially AI-generated-meaning roughly 1 in 4 posts contain substantial AI content. Philosophy follows at 23%, and Health at 22%.
The trend drops sharply in other categories. Culture contains 13% AI content, Sports 5%, Food and Drink 3%, and Music just 1%.
Readers want human voices for personal subjects
The pattern suggests readers accept AI-generated content for analytical and informational topics but prefer human authorship for subjects built around personal perspective. Restaurant reviews, music criticism, and cultural commentary skew heavily toward human writers.
Publishers using AI tend to use it extensively. Some newsletters publish entirely AI-generated content with minimal human editing, according to Pangram.
A handful of large publishers drive category-wide statistics. One AI-generated newsletter accounts for 35% of all majority-AI posts in the News category. Two Culture publications generate roughly three-quarters of AI content in that category. Just 29 Substack publications produced half of all majority-AI posts across all categories.
Success concentrates among top publishers in some fields
In Business, the top 10 publishers average 14.4% AI usage, while publishers ranked 15 and below average 0.8%. Art & Illustration, Parenting, and Sports show similar patterns, where higher-ranked publications use more AI than lower-ranked competitors.
The analysis has limitations. Some newsletters paywall portions of their content, so Pangram only processed publicly visible previews. The detection tool requires at least 50 words to determine a result and defaults toward "human" when uncertain.
What this means for writers
The findings suggest a market still values human-written content. Two-thirds of bestselling newsletters operate without detectable AI use, indicating that human-written material can drive financial success on the platform.
Writers considering AI integration should approach it strategically. Some established tech writers on Substack openly use AI for drafting and editing. Others may experiment with AI for specific tasks-like generating link roundups-while maintaining human authorship for core content.
For readers, the takeaway is straightforward: scrutiny matters most in Technology, Philosophy, Health, and Business categories, where AI content is most prevalent. Tools like Pangram's Chrome extension can identify AI-generated text across Substack, LinkedIn, X, and Reddit.
Learn more about how AI tools work in content creation through Generative AI and LLM Courses or explore practical strategies in AI for Writers.
Your membership also unlocks:
