1 in 10 top substack authors use AI to create content: report

Content generated by artificial intelligence is increasingly finding its way onto the Internet. A study conducted earlier this year found that over 40 percent of posts on the Medium publishing platform were probably created with the help of AI, while the newly created Wikipedia pages were analyzed in August 5 percent was marked as containing content generated by artificial intelligence. The new technology is also widely used by some of the world’s most prominent authors Basepopular newsletter platform with over 35 million subscribers.

Of the top 100 Substack authors, 10 percent use some form of detectable AIfirst, according to a recent report by AI detection software GPTZero reported by Wired. GPTZero came after release OpenAIChatGPT as a tool for identifying text produced by large language models (LLM). It claims to be 99 percent accurate at distinguishing AI from human-written text and 96 percent accurate at detecting whether writing contains text written by both AI and humans.

Founded in 2017, Substack has attracted a stable of well-known writers such as journalist over the years Glenn Greenwald and historian Heather Cox Richardson. GPTZero took a look at the latest content of the platform’s top 100 newsletters by pulling the most recent 25-30 posts from its feed and running through its detection model. For paid newsletters, GPTZero paid for the subscription whenever possible.

To be among the 90 percent of Substack newsletters identified as written by humans, they didn’t even have to be completely AI-free — GPTZero still awards its “Certified Human Badge” to authors who have one to two AI-generated posts. Meanwhile, according to GTPZero, 10 percent of writers were flagged as using some form of detectable AI, and 7 percent of authors were found to have significantly used AI in more than 1 in 10 posts. Nearly all of the sub-categories in this latter section boast six-figure subscriber counts and focus on topics such as sports, financial advice, and business.

How do Substack authors use artificial intelligence?

One of the seven AI-powered sub-stacks is the soccer newsletter he authored David Skillingwho also serves as the general director of the Freedom Sport sports agency. “I see AI as a support tool rather than a creator,” Skilling told the Observer, adding that AI tools have taken over the assistant role. Skilling, whose newsletter has 623,000 subscribers, uses artificial intelligence to help research stories and edit copy. He compares his use of this technology to when photographers went from developing film in darkrooms to now using “digital tools to streamline editing.”

Josh Belangerwhich chronicles the stock market for its 352,000 subscribers via its Belanger Trading newsletter, draws from LLM firms such as ChatGPT, Claude and even Elon MuskGrok to speed up your research and add color and personality to your writing. “It just helps you get a lot more things done faster,” he told the Observer, adding that he started using artificial intelligence extensively in the last six to eight months.

Other Substack authors identified in the GPTZero report as using AI say the technology augments their content rather than creating it. Subham Panda, one of the authors of Spotlight by Xartup Substack, told Wired that it uses artificial intelligence to create images and aggregate information; and Max Avery, author of the newsletter Strategic Wealth Briefing With Jake Claver, said the technology is useful for editing rough drafts.

Substack does not prohibit AI-generated content, although the platform has mechanisms to detect spam activity such as duplicate content and bot activity that often involve AI. “We do not proactively monitor or remove content based solely on its AI origins, as there are many important, constructive uses for assisted content creation,” the company said in a statement to the Observer.

GPTZero, which does not currently measure factors such as accuracy or writing quality, maintains that its AI detection capabilities increase the clarity of AI-generated text. The goal of the GPTZero report “is not to morally condemn writers who use AI,” the company said in its report, but instead “to raise awareness of the prevalence of AI-generated content, especially as the growth of AI-powered content continues unabated.”