A recent analysis has shed light on the growing usage of artificial intelligence (AI) in content creation across various online platforms, specifically focusing on the Substack newsletter service. Conducted by GPTZero, a firm known for its AI detection tools, the study highlights the reliance on AI-generated content among some of the most popular newsletters in the Substack ecosystem.
The analysis examined a sample of 100 newsletters on Substack, identifying that approximately 10 percent of these publications are likely incorporating AI-generated content to some degree. Furthermore, it noted that 7 percent of the newsletters significantly depend on AI-generated material. The findings emerged from an extensive scan of 25 to 30 posts from each selected newsletter, conducted using GPTZero’s detection technology, which assesses the likelihood that text was produced by generative AI models such as ChatGPT.
Out of the ten newsletters identified as using AI content, seven confirmed to WIRED that artificial intelligence tools play a role in their writing process. The remaining three newsletters did not respond to requests for clarification. Many publications that did provide feedback indicated that they utilise AI for various supplementary tasks, which include posting on social media, summarising content, generating images, and grammar checking, rather than relying solely on AI to compose the newsletters in full.
Notably, David Skilling, the operator behind the Original Football Substack, which has an impressive following of 630,000 subscribers, emphasised the distinction between AI-generated and AI-assisted content. He stated, “I proudly use modern tools for productivity in my businesses,” highlighting the value many creators find in leveraging AI tools to bolster their productivity.
Substack's policies regarding AI use remain ambiguous, as there is currently no explicit ban on employing generative AI for content creation on the platform. Helen Tobin, Substack’s head of communications, refrained from commenting directly on the GPTZero findings. However, she indicated that the platform has measures in place to detect and tackle inauthentic activities, such as spam and bot-generated content. These measures, however, do not focus on the AI origins of content, as Substack acknowledges that there are numerous legitimate applications for AI-assisted writing.
To contextualise the findings from Substack, GPTZero's prior assessment determined that roughly one in 20 articles on Wikipedia featured AI-produced content, a figure which is significantly lower than that of Substack newsletters. In contrast, analysis from other AI detection tools revealed that nearly 40 percent of posts on the Medium platform contained AI-generated components, although the engagement metrics for Medium’s content are often considerably lower than those seen with the top Substack newsletters.
This examination underscores the intricate relationship between emerging AI technologies and contemporary publishing practices, illustrating both the advantages and complexities that AI integration brings to content creation across various platforms.
Source: Noah Wire Services