Substack takes on “Claudefishing” as platforms step up AI transparency

Substack has launched a tool that lets readers check whether posts were written by a person or with AI assistance, becoming the latest platform to introduce new transparency measures as concerns grow over synthetic content online.

The feature, built with AI detection company Pangram, scans posts, Notes, comments and replies longer than 100 words and estimates how much text was written by hand or with AI assistance. It applies to content published from 21 July and is available on web and iOS, with Android to follow.

Substack chief executive Chris Best announced the update in a post on the website, coining the term "Claudefishing" to describe passing off AI-generated writing as human thought. The issue, he said, is not the use of AI itself, but readers being unable to tell the difference.

He cited Pangram estimates that as much as 40% of text on some social platforms is AI-generated.

"When readers have to wonder if what they're reading is real, it undermines trust in authorship and threatens the livelihood of writers," he wrote.

Alongside the detector, writers can add a "How I make this" statement explaining their creative process, scan drafts before publishing and challenge incorrect AI classifications. Those who disable detection will display an "AI detection unavailable" notice to readers.

The rollout prompted debate among users who openly use AI tools, with several arguing a detection score cannot distinguish between work thoughtfully edited by a human using AI and content generated almost entirely by a machine. Others warned that transparency could become a "purity test".

While acknowledging the limitations, Best nevertheless maintained that giving readers more context is preferable to leaving them guessing.

Substack is not the first platform to recently introduce new AI transparency measures.

Earlier this month, TikTok announced a package of updates to help people spot artificial content and reduce spam.

They included a new in-app AI learning hub and closer work with Content Credentials – a system that helps identify AI-generated content.

At the time, TikTok said AI-generated spam was "crowding out original creators" and revealed it had already labelled more than three billion videos as AI-generated.

TikTok is also testing a separate tool that allows some US creators to detect and report unauthorised AI-generated versions of their likeness, following a similar rollout by YouTube.

Donatas Smailys, chief executive of creator marketing platform Billo, said TikTok's own announcement points to the reason platforms are introducing more AI transparency measures.

"Platforms protect whatever keeps people watching," he told The Daily Influence. "When they start building detection against a format, that tells you what the format is doing to watch time."

"The evidence is in TikTok's own wording. They said AI spam is crowding out real creators. If synthetic content held people's attention, it would be earning its place in the feed and TikTok would be selling more of it."

The announcements follow a broader trend across creator platforms, with Instagram favouring original content and Spotify introducing verification badges for human artists earlier this year.