Substack’s new AI detection tool aims to bring more transparency to a platform where human‑written and AI‑generated content increasingly sit side by side. The feature, powered by Pangram, scans posts, notes, replies and comments to estimate how much text may have been produced with AI assistance.
Image: theverge.com
Overview
The Verge reports that Substack is rolling out the detector across the web and its iOS app, with an Android release coming soon. The tool does not label content as AI‑generated; instead it provides a percentage estimate, letting readers decide whether to trust the authorship of a piece.
Key points:
Powered by Pangram, an AI‑detection company.
Works on posts, notes, replies and comments longer than 100 words.
Readers access it via the three‑dot menu on any Substack item.
How the Detector Works
When a user selects “Scan for AI text,” Pangram analyses the linguistic patterns in the selected text and returns a score that reflects the likelihood of AI involvement. Substack co‑founder Chris Best emphasizes that the tool only detects whether AI was used to create the text, not the quality of any human editing that may have followed.
Tip: Writers can run the detector on their own drafts and, if they believe the result is inaccurate, submit a report to help improve the model.
Why This Matters
Best describes the core issue as a mismatch between reader expectations and reality—what he calls “Claudefishing.” When audiences unknowingly invest attention in content that lacks genuine human thought, trust erodes and the livelihood of thoughtful creators suffers. By surfacing AI usage, Substack hopes to restore confidence in authorship.
Note: The detector does not judge whether AI‑assisted work is valuable; it merely signals the presence of AI‑generated text.
What to Do Next
Readers should look for the three‑dot menu on any Substack post longer than 100 words and select “Scan for AI text” to view the estimate. Creators who want to be transparent can add a “How I make this” note alongside their work, explaining their use of AI tools.
Potential Impact on Creators and Readers
If widely adopted, the detector could shift platform dynamics: readers may gravitate toward verified human‑written newsletters, while creators who use AI responsibly might highlight their process to stand out. Conversely, some fear the tool could stigmatize AI‑assisted work, even when it is carefully curated.
Limitations and Considerations
Pangram’s technology cannot assess the depth of human editing or the intent behind AI use. Best notes that a high score does not automatically mean low quality, and a low score does not guarantee thoughtful human authorship. The tool is best used as one signal among many when evaluating content.
Want more practical guides? Explore more Ayxworks insights and save this article for later.
Substack adds an AI detector to help spot blogs written by no one The new tool can pick up on ‘Claudefishing’ across posts, comments, notes, and replies.
by Emma Roth Jul 21, 2026, 7:22 PM UTC Image: The Verge Emma Roth is a news writer who covers the streaming wars, consumer tech, crypto, social media, and much more.
Previously, she was a writer and editor at MUO.
The tool is powered by an AI detection company called Pangram and is rolling out across the web and the iOS app, with an Android launch coming “soon.” Readers can analyze content longer than 100 words by choosing the “Scan for AI text” option from the three-dot menu in the top-right corner of a post.
Image: Substack Substack co-founder and CEO Chris Best says the Pangram integration is meant to increase transparency on the platform as AI becomes more prevalent on social media.
“The core problem is not people using AI, or the quality of its output,” Best writes.
FAQ
What is Substack’s new AI detector?
It is a Pangram‑powered feature that scans Substack posts, notes, replies and comments longer than 100 words to estimate the proportion of text that may have been AI‑generated or AI‑assisted.
How accurate is the Pangram‑powered detection tool?
The Verge article does not provide a specific accuracy percentage; it describes the tool as an estimate that helps readers gauge AI involvement rather than a definitive verdict.
Can creators use the detector on their own drafts?
Yes. Writers can run the scan on drafts before publishing and, if they believe the result is incorrect, submit a report to help refine the detection model.
What does “Claudefishing” mean in this context?
Co‑founder Chris Best coined the term to describe the situation where readers unknowingly devote attention to content that lacks genuine human thought, undermining trust in authorship.
Is the detector available on mobile apps?
The feature is rolling out to the web and iOS app now, with an Android launch described as “coming soon” in the source article.
Will Substack penalize content flagged as AI‑generated?
Substack has not announced any penalties or ranking changes based on the detector’s output; the tool is intended solely for transparency and informing reader choice.
External Sources and Further Reading
