Blab Tech
Substack Launches AI-Detection Tool to Counter AI Slop on Its Platform
When Substack rolled out its new AI‑detection feature on Tuesday, it didn’t merely add another app function—it gave readers a way to see who really wrote the words they read.
The subscription‑based newsletter platform announced that the new tool, built in partnership with Pangram Labs, can scan any note, reply, comment or post and estimate how much of the text was generated by a human versus an AI. The feature is now live on both the web and mobile apps.
The move follows a growing alarm that the internet is awash with low‑quality, AI‑generated content—colloquially known as “AI slop.” Substack’s chief executive, Chris Best, made clear that the company is not against AI tools in general. He noted that the platform already employs AI internally for software development, research, and product features such as clipping and translation.
Instead, Best said the goal is to empower users to identify AI‑generated material. "We’re not against people using AI to assist their work," he wrote. "We think people should be free to choose which tools they use to express themselves." He added that Substack is exploring future options that would let readers filter AI‑generated content from their feeds.
Pangram’s detector is described as “industry‑leading” and “imperfect.” According to Pangram’s website, the system uses natural‑language‑processing models trained on a large dataset of human and AI writing. It can distinguish between fully AI‑generated text and content that has been partially edited by a human. The detector accepts up to 5,000 words and supports TXT, DOCX and PDF uploads.
Substack’s implementation is currently limited to posts longer than 100 words. Users can launch a scan by selecting the three‑dot menu in the top‑right corner of a post and choosing "Scan for AI text." The feature does not apply to video or audio posts, nor to content hosted on custom domains or the on.substack.com subdomain.
The announcement arrives amid a broader trend of platforms adding AI‑detection tools. Earlier this week, The Verge and Engadget reported on Substack’s rollout, noting that the company is not enforcing a policy that bans AI‑generated content. A 2025 E‑Week report found that roughly 10% of Substack’s largest newsletters use AI tools, but the platform has no official stance on the matter.
Substack’s decision reflects growing awareness that AI‑generated text can be misleading or low quality. The term "AI slop" describes content that is fluent yet shallow, often produced in bulk to capture clicks or ad revenue. A 2026 Columbia University study labeled AI slop as "low‑quality, templated AI content that says nothing."
Industry observers see AI detection as a means to restore trust in online publishing. By giving readers a transparent assessment of authorship, platforms can help curb misinformation and preserve the value of human‑written journalism.
In a statement, Best reiterated that the company’s goal is not to stifle AI innovation but to provide tools that let users make informed choices. "We don’t want to wait until our app turns into LinkedIn before we start to learn and make progress," he wrote, referencing concerns that a platform could become a hub for AI‑generated content.
The rollout is part of Substack’s broader strategy to address consumer anxiety about AI. While the platform remains open to AI‑assisted writing, it now offers a mechanism for readers to identify—and if they wish, filter out—AI‑generated material.
At this time, Substack has not announced any policy changes that would restrict the use of AI tools by writers. The company’s focus appears to be on transparency and user choice rather than censorship.
In the coming months, Substack may expand the detector’s capabilities and explore additional features that allow readers to curate their content streams. The partnership with Pangram demonstrates a willingness to engage with third‑party solutions to address the challenges posed by AI slop.
Overall, the launch signals that even platforms built around human authorship are recognizing the need for tools that help users navigate an increasingly AI‑rich digital landscape.
The subscription‑based newsletter platform announced that the new tool, built in partnership with Pangram Labs, can scan any note, reply, comment or post and estimate how much of the text was generated by a human versus an AI. The feature is now live on both the web and mobile apps.
The move follows a growing alarm that the internet is awash with low‑quality, AI‑generated content—colloquially known as “AI slop.” Substack’s chief executive, Chris Best, made clear that the company is not against AI tools in general. He noted that the platform already employs AI internally for software development, research, and product features such as clipping and translation.
Instead, Best said the goal is to empower users to identify AI‑generated material. "We’re not against people using AI to assist their work," he wrote. "We think people should be free to choose which tools they use to express themselves." He added that Substack is exploring future options that would let readers filter AI‑generated content from their feeds.
Pangram’s detector is described as “industry‑leading” and “imperfect.” According to Pangram’s website, the system uses natural‑language‑processing models trained on a large dataset of human and AI writing. It can distinguish between fully AI‑generated text and content that has been partially edited by a human. The detector accepts up to 5,000 words and supports TXT, DOCX and PDF uploads.
Substack’s implementation is currently limited to posts longer than 100 words. Users can launch a scan by selecting the three‑dot menu in the top‑right corner of a post and choosing "Scan for AI text." The feature does not apply to video or audio posts, nor to content hosted on custom domains or the on.substack.com subdomain.
The announcement arrives amid a broader trend of platforms adding AI‑detection tools. Earlier this week, The Verge and Engadget reported on Substack’s rollout, noting that the company is not enforcing a policy that bans AI‑generated content. A 2025 E‑Week report found that roughly 10% of Substack’s largest newsletters use AI tools, but the platform has no official stance on the matter.
Substack’s decision reflects growing awareness that AI‑generated text can be misleading or low quality. The term "AI slop" describes content that is fluent yet shallow, often produced in bulk to capture clicks or ad revenue. A 2026 Columbia University study labeled AI slop as "low‑quality, templated AI content that says nothing."
Industry observers see AI detection as a means to restore trust in online publishing. By giving readers a transparent assessment of authorship, platforms can help curb misinformation and preserve the value of human‑written journalism.
In a statement, Best reiterated that the company’s goal is not to stifle AI innovation but to provide tools that let users make informed choices. "We don’t want to wait until our app turns into LinkedIn before we start to learn and make progress," he wrote, referencing concerns that a platform could become a hub for AI‑generated content.
The rollout is part of Substack’s broader strategy to address consumer anxiety about AI. While the platform remains open to AI‑assisted writing, it now offers a mechanism for readers to identify—and if they wish, filter out—AI‑generated material.
At this time, Substack has not announced any policy changes that would restrict the use of AI tools by writers. The company’s focus appears to be on transparency and user choice rather than censorship.
In the coming months, Substack may expand the detector’s capabilities and explore additional features that allow readers to curate their content streams. The partnership with Pangram demonstrates a willingness to engage with third‑party solutions to address the challenges posed by AI slop.
Overall, the launch signals that even platforms built around human authorship are recognizing the need for tools that help users navigate an increasingly AI‑rich digital landscape.
