Substack launches AI detection tool to identify AI-generated content and combat Claudefishing

Substack Introduces AI Detection Feature to Identify AI-Generated Content

Substack is rolling out an AI detection tool that allows readers to scan for AI text across the platform, marking a notable shift in how content platforms address transparency in authorship. The feature, powered by AI detection company Pangram, can analyze posts, notes, replies, and comments longer than 100 words to provide an estimate of how much text could be AI-generated 1

1

. Readers can access this AI detection feature by selecting the "Scan for AI text" option from the three-dot menu in the top-right corner of any post 1

1

.

The tool is currently available on web and iOS, with Android support expected to arrive soon 2

2

. This deployment signals Substack's commitment to helping readers understand what they're consuming on the platform as AI-generated content becomes increasingly prevalent across social media.

Addressing Claudefishing and Reader Expectations

Substack co-founder and CEO Chris Best framed the initiative around a concept he calls "Claudefishing"—a term describing situations where readers unwittingly invest their attention in content with no human thought behind it. "The core problem is not people using AI, or the quality of its output," Best writes. "The problem is when there is a mismatch between a reader's expectation and reality, especially when they unwittingly invest their attention in something with no human thought on the other end" 1

1

.

This focus on reader expectations reflects growing concerns about authenticity in digital publishing. Best emphasized that when readers have to question whether content is real, it undermines trust in authorship and threatens writers' livelihoods—including those who use AI tools thoughtfully 1

1

. The announcement also included a pointed reference to LinkedIn, highlighting concerns about AI-generated content on that service 2

2

.

New Tools for Writers to Explain Their Writing Process

Alongside the AI detection tool, Substack is introducing a "How I make this" statement feature that allows creators to explain their writing process to readers. Writers can scan their own drafts with Pangram before publishing and will have an option to report inaccurate results 1

1

. This dual approach addresses both reader concerns and creator transparency, giving writers a way to proactively communicate how they use AI tools in their work.

Best acknowledged limitations in what Pangram can detect, noting that the tool "can only detect whether AI was used to make the text, not whether great human care went into creating it, nor whether AI tools were used as a source" 1

1

. This caveat matters because many writers use AI as a research assistant or editing tool rather than for generating entire pieces.

Questions About Accuracy and Long-Term Implications

While Substack positions this as a transparency measure, the accuracy of AI detection remains uncertain. Even the best tools for identifying AI-generated content cannot guarantee correct assessments, and The Atlantic has examined Pangram's accuracy specifically, finding these tools are far from perfect 2

2

. This raises questions about potential false positives that could unfairly flag human-written content or false negatives that miss AI-generated material.

Substack acknowledged these limitations and hinted at additional features it is considering around AI content and preferences 2

2

. The platform's stance, as Best summarized, is that "people should know what they're getting" 2

2

. He warned that "platforms that reward fakeness will create a race to the bottom" 1

1

, suggesting this move is part of a broader strategy to differentiate Substack from other content platforms where AI-generated material proliferates unchecked. As AI tools become more sophisticated, the challenge of maintaining transparency in authorship will likely intensify, making Substack's approach a test case for how platforms balance creator freedom with reader trust.🟡 untrained_content=🟡### Substack Introduces AI Detection Feature to Identify AI-Generated Content

Substack is rolling out an AI detection tool that allows readers to scan for AI text across the platform, marking a notable shift in how content platforms address transparency in authorship. The feature, powered by AI detection company Pangram, can analyze posts, notes, replies, and comments longer than 100 words to provide an estimate of how much text could be AI-generated 1

1

. Readers can access this AI detection feature by selecting the "Scan for AI text" option from the three-dot menu in the top-right corner of any post 1

1

.

The tool is currently available on web and iOS, with Android support expected to arrive soon 2

2

. This deployment signals Substack's commitment to helping readers understand what they're consuming on the platform as AI-generated content becomes increasingly prevalent across social media.

Addressing Claudefishing and Reader Expectations

Substack co-founder and CEO Chris Best framed the initiative around a concept he calls "Claudefishing"—a term describing situations where readers unwittingly invest their attention in content with no human thought behind it. "The core problem is not people using AI, or the quality of its output," Best writes. "The problem is when there is a mismatch between a reader's expectation and reality, especially when they unwittingly invest their attention in something with no human thought on the other end" 1

1

.

This focus on reader expectations reflects growing concerns about authenticity in digital publishing. Best emphasized that when readers have to question whether content is real, it undermines trust in authorship and threatens writers' livelihoods—including those who use AI tools thoughtfully 1

1

. The announcement also included a pointed reference to LinkedIn, highlighting concerns about AI-generated content on that service 2

2

.

New Tools for Writers to Explain Their Writing Process

Alongside the AI detection tool, Substack is introducing a "How I make this" statement feature that allows creators to explain their writing process to readers. Writers can scan their own drafts with Pangram before publishing and will have an option to report inaccurate results 1

1

. This dual approach addresses both reader concerns and creator transparency, giving writers a way to proactively communicate how they use AI tools in their work.

Best acknowledged limitations in what Pangram can detect, noting that the tool "can only detect whether AI was used to make the text, not whether great human care went into creating it, nor whether AI tools were used as a source" 1

1

. This caveat matters because many writers use AI as a research assistant or editing tool rather than for generating entire pieces.

Questions About Accuracy and Long-Term Implications

While Substack positions this as a transparency measure, the accuracy of AI detection remains uncertain. Even the best tools for identifying AI-generated content cannot guarantee correct assessments, and The Atlantic has examined Pangram's accuracy specifically, finding these tools are far from perfect 2

2

. This raises questions about potential false positives that could unfairly flag human-written content or false negatives that miss AI-generated material.

Substack acknowledged these limitations and hinted at additional features it is considering around AI content and preferences 2

2

. The platform's stance, as Best summarized, is that "people should know what they're getting" 2

2

. He warned that "platforms that reward fakeness will create a race to the bottom" 1

1

, suggesting this move is part of a broader strategy to differentiate Substack from other content platforms where AI-generated material proliferates unchecked. As AI tools become more sophisticated, the challenge of maintaining transparency in authorship will likely intensify, making Substack's approach a test case for how platforms balance creator freedom with reader trust.