eWEEK content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.
A new analysis found that 10 percent of the biggest newsletters on the Substack publishing platform likely use AI-generated content in some capacity, and 7 percent of the newsletters “significantly rely” on AI content. The analysis was conducted by GPTZero, whose AI detector tool can determine the likelihood of text having been created by a generative AI. The results were published exclusively in WIRED.
GPTZero examined the 100 most popular newsletters on the Substack platform, scanning 25 to 30 posts from each newsletter with its AI detection tool. GPTZero found that 10 of the newsletters likely use AI-generated content made by ChatGPT and other platforms, with seven of those publications heavily relying on it.
WIRED reached out to all seven newsletters. According to the magazine, four of them confirmed that artificial intelligence tools are part of their writing process. The remaining three did not respond to requests for comment. However, most of the publications that did respond said they use AI tools for posting to social media, creating summaries, making images, checking posts for grammar, and other related tasks rather than fully generating the Substack newsletter content.
“I proudly use modern tools for productivity in my businesses,” said David Skilling, who runs the Original Football Substack, which boasts 630,000 subscribers. “AI-detection tools may detect the use of AI, but there’s a huge difference between AI-generated and AI-assisted.”
Substack does not have an official policy about the use of AI on its platform, so nothing prohibits writers from using generative AI to make content for Substack. Helen Tobin, Substack’s head of communications, declined to comment specifically on GPTZero’s results.
“We have several mechanisms in place to detect and mitigate inauthentic or coordinated spam activities, such as copypasta, duplicate content, SEO spam, phishing, and bot activity—many of which can involve AI-generated content,” she told WIRED in an email. “However, we don’t proactively monitor or remove content solely based on its AI origins, as there are numerous valid, constructive applications for AI-assisted content creation.”
To put the Substack AI news in perspective, another analysis by GPTZero found that about one in 20 Wikipedia articles use AI-created content—about half the rate of Substack’s results. However, additional analysis from two other AI detection tools found that nearly 40 percent of posts on the Medium publishing platform contain AI content. The difference is that most of those Medium posts get very little engagement, while the Substack perpetrators represent some of the most popular newsletters on the platform.