Substack rolls out AI detection tool to flag “Claudefishing” on its platform

Substack rolls out AI detection tool to flag “Claudefishing” on its platform

N
News Editor
2026-07-22 09:03:51
Substack has introduced a new AI detection feature aimed at giving readers more visibility into how much of a post may have been generated by artificial intelligence. The company framed the move as a response to what it calls “Claudefishing,” a term derived from Anthropic’s Claude model and used to describe cases where creators rely heavily on Claude or other large language models while presenting the work as if it were entirely their own. Substack said it is not banning AI use. The platform described several uses as acceptable, including outlining articles, correcting grammar, translating text, polishing language, and helping with research. Its concern is with posts that are largely produced by AI but are still packaged as original personal commentary, leaving readers unable to tell how much thinking and writing came from the author. The detection system is powered by AI detection company Pangram. It will analyze posts, comments, Notes, and replies, and estimate what share of a piece is likely AI-written rather than assigning a simple yes-or-no label. Substack also said no AI detector is 100% accurate, and that Pangram’s results are meant as a reference rather than a final judgment.
SubstackAI detectionClaudefishingPangramClaudeGenerative AIContent platforms

Substack has launched a new AI detection feature designed to show readers how much of the content they are reading may have been generated by AI. The company said the tool is meant to address what it calls “Claudefishing.”

What Substack means by “Claudefishing”

The term comes from Anthropic’s AI model Claude and is modeled on the idea of catfishing, or pretending to be someone else online. Substack defines it as a situation where a creator leans heavily on Claude or another large language model to produce an article while giving readers the impression that the piece was written entirely by the author.

In Substack’s view, the core issue is not AI itself but the lack of transparency. In its announcement, the platform said readers should have the right to know whether what they are reading is human-created or AI-generated.

AI use is allowed, but disclosure matters

Substack said it is not banning AI. The company described AI as an important tool for many creators and listed several acceptable uses: generating an article outline, fixing grammar, translating content, polishing wording, and assisting with research.

The behavior it wants to curb is different. Substack said the problem arises when nearly an entire piece is produced by AI and then published as if it were a writer’s original personal perspective, making it hard for readers to judge how much human thought and creative work actually went into it.

Pangram powers the detection system

The new feature is supported by AI detection company Pangram. The system will analyze posts, comments, Notes, and replies, and estimate what proportion of the text may have been written by AI instead of simply labeling content as “AI” or “not AI.”

Substack also acknowledged that no AI detection tool is 100% accurate. It said Pangram’s output should be treated as a reference, not a final ruling. The platform added that AI detection can only assess whether text shows signs of AI generation; it cannot measure the quality of the content or determine how much human thinking the author contributed.

This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan.
500

Disclaimer:

The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.

Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.