Substack, the popular newsletter and publishing platform, has taken a notable step toward content transparency by partnering with AI-detection firm Pangram. The new feature enables readers to check whether a post, note, or reply they are viewing was written by a human or generated by a large language model. CEO Chris Best announced the tools in a post titled “Against Claudefishing,” a term he coined to describe content that heavily relies on AI while pretending to be human-written.
How the Scanning Tool Works
The scanning tool can analyze any text longer than one hundred words that was published on or after July 21. The result of each scan is private and visible only to the person who requested it. Best acknowledged that Pangram’s detection is not perfect but pointed to independent evaluations that credit the tool with a high degree of accuracy. The feature is currently live on the web version and iOS app, with an Android release expected later.
Substack’s decision to implement this tool comes at a time when the problem of AI-generated content is spreading rapidly across social platforms. Best cited a recent Pangram estimate that as much as forty percent of posts on some platforms are now fully AI-generated. This has raised concerns about the authenticity of online discourse and the erosion of trust between creators and their audiences.
Writers Can Run the Same Scan Before They Publish
In addition to the reader-facing tool, Substack is introducing a “How I make this” statement that allows writers to describe their creative process and set expectations for their audience. Writers also have the option to run Pangram’s scan on their drafts before publishing, giving them a chance to adjust their work if they feel it appears too machine-like. Furthermore, authors can report and request the removal of scans on their published work if they believe the result is erroneous. Best indicated that future updates could include AI preferences within Reply Rules and reader-side controls for recommendations.
This transparency feature is rare in the tech industry, where most platforms rely on automated systems to filter content rather than giving users direct visibility. For readers, it is a practical button to spot what some call “AI slop” — generic, often repetitive content generated by chatbots. Substack is not alone in addressing this issue; LinkedIn has started reducing the reach of AI-generated posts and comments, and Meta is quietly testing its own AI detection tool, though it has yet to roll out widely. The effectiveness of these tools in keeping pace with rapidly evolving language models remains an open question.
Background on Substack’s Approach to Content
Substack was founded in 2017 as a platform that enabled individual writers to publish newsletters and monetize their work through subscriptions. Over the years, it has expanded to include podcasts, video, and community features. The platform has always emphasized the importance of independent writing and direct connections between creators and readers. The introduction of AI detection aligns with this ethos, aiming to preserve the authenticity that many subscribers value.
The rise of generative AI models like ChatGPT, Claude, and Gemini has made it increasingly difficult for readers to distinguish between human and machine writing. Many platforms have struggled to moderate AI-generated content without infringing on legitimate use cases, such as writers using AI for editing or brainstorming. Substack’s partnership with Pangram is an attempt to strike a balance, giving both readers and writers tools to navigate this new landscape.
Industry Context and Future Implications
The move also reflects broader industry trends. AI detection is a rapidly developing field, with companies like Originality.ai, GPTZero, and Copyleaks competing for market share. Pangram, a relatively new entrant, has positioned itself as a specialist in content authenticity. Substack’s choice of partner may signal a preference for a tool that focuses on contextual accuracy rather than simply flagging statistical anomalies.
Critics argue that no AI detector can be 100% accurate, and false positives could harm legitimate writers who use AI for grammar or style suggestions. Substack acknowledges this by allowing writers to dispute scans. The platform also emphasizes that the scan results are not used for punitive measures, such as demonetization or suspension, at least for now. Instead, the goal is to inform readers and encourage honest disclosure.
The long-term impact of these tools on Substack’s ecosystem remains to be seen. If widely adopted, they could set a new standard for transparency in online publishing. However, adoption depends on user trust and the tool’s perceived reliability. As generative AI continues to improve, detection tools may need constant updates to remain effective, creating an arms race between AI content generators and detectors.
For now, Substack provides a concrete option for readers who want to verify the origin of the text they engage with. Whether other platforms will follow suit is unclear, but the conversation about AI transparency in journalism and content creation is only beginning.
Source: Digital Trends News