In both the encyclical and a speech delivered the same day to Catholic Church officials, Leo said he believed AI "must" be and "demands" to be "disarmed," saying leaders should direct AI "toward the common good."
Shortly thereafter, online users and some news media outlets published a claim suggesting Leo himself may have used artificial intelligence to assist in writing about the dangers of AI.
In sum, we examined evidence that lent some credibility to the claim but did not find conclusive proof that Leo, Vatican officials or anyone else used AI to create the encyclical. As such, the status of this claim is unproven and we didn't add a fact-check rating to this article.
Snopes contacted Matteo Bruni, the director of the Holy See press office, to ask for comment on this claim and will update this article if we receive further information. We also emailed or privately messaged The Verge, a Substack writer named Linch Zhang and the AI detection platform Pangram.
The origins of the rumor
On May 26, tech news website The Verge published a brief article with the headline, "Did the Pope use AI to write about the dangers of AI?" The Verge's X post (archived) promoting its article received millions of views.
People online discussed the subject on Bluesky (archived), Facebook (archived), Instagram (archived), LinkedIn (archived) and Reddit (archived), as well as other social platforms. Other news outlets subsequently reported similar stories.
The Verge's writers cited two sources from the LessWrong community blog website, primarily a lengthy post by Substack writer Linch Zhang titled "Claude, Author of the Humanitas." The post's subhead read, "Evidence that the first papal encyclical on AI was substantially written by AI."
Zhang's post laid out evidence intended to lend credibility to the claim that whoever wrote or co-wrote the encyclical relied to some extent on the AI research company Anthropic's Claude chatbot for assistance. Zhang highlighted his own experience visually detecting potential AI-written text, as well as some of the textual clues he found, such as the repeated use of certain punctuation marks (for example, em dashes), words (for example, "genuinely") and grammatical conventions (for example, tricolons). He also documented evidence of AI use, though only in parts of the encyclical, using the AI detection platform Pangram.
Zhang argued that "the consilience of evidence across multiple angles and sources is in my opinion very hard to dismiss collectively," adding, "Any individual method might be flawed, but I believe the consilience of evidence is very strongly suggestive, perhaps even overwhelming."
Via a private message, we asked Zhang to clarify whether he believed he had amassed conclusive evidence that someone used AI to create the encyclical. Zhang responded, in part, "Without explicit confirmation from the Vatican or a confession it's essentially impossible to definitively conclude AI usage. All the evidence is statistical."
Correspondence
Pangram spokesperson Alex Roitman told us via email the company backs Zhang's data regarding the platform's results, including scanning all the encyclical's roughly 40,000 words (emphasis ours):
We stand behind the scans in this instance. Because the scans show that there was periodic AI use in the text, we would say this is mixed content. A full scan shows majority human writing with some parts (8 segments) are AI-assisted or fully AI-generated text. The length of the text also increases the likelihood that the scan is accurate.
Roitman shared Pangram's scan results, which analyzed the full text as "mostly human written" (94% human, 4% AI-generated and 2% AI-assisted). The chart displayed "8 AI segments in total" — a sporadic pattern rather than a consistent one, as Zhang would note.
(Pangram)
In an X post, Zhang corresponded with (archived) Christopher Hale, a Substack writer for the Letters from Leo blog. Hale had remarked (archived) of The Verge's article talking about AI usage for Leo's encyclical, "I can confirm that this is 100% false. Not only was it written by hand, but the first drafts were literally written on paper with a pen."
Hale told us via WhatsApp that the Holy See writes an encyclical's first draft with "pen/pad sessions" and "through different dicasteries in the Vatican." (Dicasteries are administrative departments.) We asked him for evidence and will update this story if we receive more details.
Zhang responded to Hale's X post, in part pointing out The Verge did not accurately include in its reporting that Zhang said three times he believed cardinals, and "probably" not Leo himself, contributed to the encyclical using some AI. We emailed The Verge to ask about that omission and did not receive a response by publication time.
"Different sections of the encyclical have very different rates of apparent AI usage," Zhang's post said. "This indicates to me that some cardinals used AI assistance for this encyclical and many (probably including Pope Leo himself) don't." He added, "My tentative hypothesis is that Pope Leo does not approve of the AI usage in encyclicals, and plausibly was not even aware of significant AI usage in his own encyclical! Quite unfortunate if true."
Zhang later told us, "I've since learned after writing my post that it's typically curial staff at the Vatican who specialize in writing who do most of the drafting, rather than cardinals (who are also very busy)." We weren't able to confirm that exact information, but the Encyclopedia of Global Religion describes papal encyclicals as "often the result of collaborative work."
In The Verge, the writers accurately reported, "AI detection isn't foolproof. Different AI detectors can display different results, and even when there's consensus there's no guarantee they're correct." The writers also said AI researchers generally respect Pangram.
Roitman, the Pangram spokesperson, cited third-party evaluations documenting a false-positive rate of below 0.05% for its detection tool, as well as internal data documenting a false-positive rate of 0.02% in academic writing.
In late May 2026, The Atlantic reported about the dangers of placing too much trust in Pangram and mentioned Leo's encyclical as a prominent example of a text identified as AI-generated. The reporting said, in part, "While Pangram is accumulating the power to end reputations and careers, the tool does make mistakes, perhaps to a greater extent than is currently understood. In turn, AI accusations could very quickly spiral into a witch hunt."
Analyzing past encyclicals
Using Pangram, Zhang scanned the lengthy beginnings of past encyclicals — including those written by popes Francis, Benedict XVI and John Paul II — for signs of AI use. Pangram determined a 100% "human written" score — meaning no detected AI — in all those cases. Those tests were conducted mainly to gauge Pangram's reliability rather than the integrity of the writings themselves, seeing as AI tools didn't exist at those earlier times.
Snopes performed the same tests with encyclicals from Francis, Benedict and John Paul II, scanning the most recent five documents from the latter. Pangram reported a 100% "human written" score for all of them, confirming Zhang's data.
Zhang's research also included a scan of Leo's speech on the day "Magnifica Humanitas"' was published, with Pangram reporting no AI content. We got the same results.