Analyses Flag Parts of Papal Encyclical as AI-Written

Analyses published and discussed online indicate that portions of Magnifica Humanitas, Pope Leo XIV's encyclical on artificial intelligence, were flagged as AI-generated by automated detectors. Reporting in The Verge cites a LessWrong post by Linch Zhang that used the detector Pangram, finding certain paragraphs assessed as between 40 percent and 100 percent AI-written, per Zhang as reported by The Verge. A separate user reportedly found 62 percent of the first chapter flagged, and The Verge reports it ran roughly 2,000 words of the document through Pangram, which returned an aggregate estimate of 46 percent AI-written.

Vatican News published an overview of Magnifica Humanitas and reports the encyclical was signed on May 15 and published on May 25. Time describes the document as a roughly 42,300-word encyclical that frames AI as a sweeping social and moral challenge and notes the pope presented the text at the Vatican alongside Christopher Olah, a co-founder of Anthropic.

Automated detectors like Pangram use stylistic and statistical signals to estimate likelihood of AI generation, and their outputs are probabilistic rather than definitive. The Verge reports that Pangram provided a quoted estimate that its false positive rate of labeling human-written work as AI-generated is "to be approximately 1 in 10,000." The Verge also underscores that different detectors can give different results and that sections of the encyclical were flagged as effectively human-written by the same tools.

The intersection of a high-profile religious document and AI-authorship claims raises two recurring industry themes: the limits of current attribution technology, and the reputational implications when public-trust institutions publish text that third parties question. The encyclical itself addresses concentration of power, labor, and dignity in the age of AI, themes covered in reporting by The Washington Post, Time, and Vatican News.

• •Observers will watch for any statement from the Holy See clarifying authorship or editorial process for Magnifica Humanitas.
• •Independent forensic analyses that publish methodology and raw comparisons will be important to corroborate or refute detector-based claims.
• •Reporting that traces manuscript drafts, editorial chains, or the involvement of outside researchers or consultants would materially change the evidentiary picture.

Stylometric and detector-based assessments can be confounded by multiple causes-drafting workflows that include outside contributors, heavily edited machine output, translators, or editorial revisions-so a detector flag alone does not establish how a text was composed. The Verge highlights that some paragraphs were rated "essentially 0% AI" by the same detector, illustrating intra-document variation.

The circulating detector results are newsworthy because they touch on authenticity and trust in public moral claims, but they do not, by themselves, prove how the encyclical was produced. Practitioners should treat single-detector outputs as suggestive and seek multi-method forensic work before drawing firm conclusions.