The Verge has raised an eyebrow-raising question about Pope Leo XIV's recent encyclical, Magnifica Humanitas, which addresses the ethical and human implications of artificial intelligence. An analysis by Linch Zhang, posted on the forum LessWrong, found that certain paragraphs of the document registered between 40 percent and 100 percent likely to have been written by AI, according to popular AI detection tools. The irony of a papal document warning about AI's impact on humanity potentially being drafted, at least in part, by the very technology it cautions against has not been lost on observers.
To understand why this matters, it helps to know something about how encyclicals are actually made. A papal encyclical is one of the most authoritative documents a pope can issue, carrying significant doctrinal and moral weight across the global Catholic Church and beyond. But these are rarely solo efforts. They typically emerge from a collaborative process involving Vatican theologians, curial officials, and subject-matter experts who research, draft, and revise over extended periods. The pope's role is often to shape, approve, and lend his voice and authority to the final product rather than to sit down and type every word himself. Ghost-writing, or something functionally close to it, has always been part of the encyclical tradition. That historical context does not make the AI question irrelevant, but it does complicate any simple narrative of hypocrisy.
What is genuinely new here is the tool being introduced into that drafting process, or at least suspected of having been introduced. The Catholic Church has been watching artificial intelligence with unusual attentiveness. The Vatican's Pontifical Academy for Life has engaged with AI ethics for several years, and the previous pontificate engaged with technology researchers and ethicists in ways that would have seemed implausible a generation ago. Magnifica Humanitas, as reported by The Verge, appears to continue that engagement, treating AI not merely as a technical concern but as a civilizational one. The question of who or what actually wrote the document does not undermine that concern — it arguably deepens it.
AI detection tools are also deeply imperfect instruments, and that caveat deserves serious weight here. These tools, including the one referenced in Zhang's LessWrong analysis, work by measuring statistical patterns in language. They tend to flag writing that is formal, structured, and syntactically smooth — qualities that Church documents have cultivated for centuries entirely without machine assistance. Latin-derived ecclesiastical prose, translated and polished by committees of scholars, can read to a language model's pattern-recognition systems as suspiciously similar to AI output precisely because both draw on the same well of formal register and careful construction. The likely reading is that the detection scores reflect a genuine ambiguity, not a confirmation.
Nevertheless, the possibility cannot simply be dismissed. If the Vatican did use a large language model as a drafting aid, even in a limited capacity, it would mark a significant institutional moment. Governments, law firms, medical institutions, and newsrooms have all been grappling with where AI assistance ends and AI authorship begins. The Catholic Church would be navigating the same question as any other large organization producing complex documents under time pressure — which is, in its own way, a remarkable sign of how normalized these tools have become. The story would then be less about hypocrisy and more about the universality of the adoption curve.
The consequences ripple outward in a few directions. For the Church, the reputational stakes are real but manageable, provided the Vatican is transparent about its drafting processes. Catholic institutions are not immune to credibility questions, and a document warning about AI's dehumanizing potential while relying on it for prose would invite sustained criticism from both secular commentators and traditionalists within the Church itself. For the broader AI ethics conversation, the episode is a useful reminder that the technology does not respect the boundaries of the debates being held about it — it is already inside those debates, shaping the language in which they are conducted. And for AI detection tools, this episode underscores their continued unreliability as arbiters of anything consequential. Scores in the range Zhang reported cannot and should not be treated as proof.
What to watch for next is threefold. The Vatican may respond directly to the question of how Magnifica Humanitas was drafted, and that response, whatever it says, will itself become part of the encyclical's legacy. Researchers and journalists will likely scrutinize the document further using multiple detection methodologies, though none will be definitive. And the broader question of whether religious institutions will develop formal disclosure norms around AI-assisted writing — the way some academic journals have begun to require — may now have a high-profile test case to accelerate that conversation.
