Tool Flags Pope Leo XIV’s X Posts as Likely AI-Written

Pangram Labs’ Chrome extension recently flagged several posts on Pope Leo XIV’s official X account as likely AI-generated. The flagged messages included warnings about artificial intelligence’s effects on human thought and social structures.

One flagged post read: “When simulation becomes the norm, it weakens the human capacity for discernment.”

Pangram says its updated browser tool scans product reviews, social posts and news articles in real time, comparing text to patterns its model associates with AI generation. The company reported that the pope’s account contained multiple messages that matched those patterns.

Max Spero, chief executive of Pangram Labs, described the findings as indicating the pope does not personally run the account and that the Vatican’s social media staff appear to be using some degree of AI.

Pangram reports its detection model has a 99.98% accuracy rate and a false positive rate of one in 10,000. A University of Chicago evaluation in December 2025 ranked Pangram highest among the tools it tested and found a near-zero false positive rate for the company’s model. Pangram Labs and Vatican representatives did not respond to requests for comment.

The extension also flagged posts from verified influencers on X, several popular Substack writers and a message from Apple’s CEO marking the company’s 50th anniversary. Pangram launched the public-facing browser tool amid growing attention to AI-generated content and says the extension aims to help users identify synthetic text online.

AI-detection tools remain controversial. In 2024, a widely used detector labeled the U.S. Declaration of Independence as more than 97% likely to be AI-generated, an error highlighted by Christopher Penn, chief data scientist at Trust Insight. Penn has warned that such errors have led to students being accused of using AI and facing academic penalties.

Researchers at Stanford University, Imperial College London and the Internet Archive estimated that roughly 35% of newly published websites contained AI-generated or AI-assisted text by mid-2025. The University of Chicago evaluation noted that no detector is perfect and advised organizations to weigh the risk of AI misuse against the possibility of false accusations when deciding how to use these tools.