New research finds top AI models show bias toward Catholicism
A consortium of religious universities tested 14 artificial intelligence systems and found they consistently favored Catholicism while disfavoring several other faith traditions.
A new study from a coalition of religiously affiliated universities has found that most popular artificial intelligence models show a measurable bias toward Catholicism and against several other religious traditions, particularly when asked about religious conversion.
The findings were released Tuesday alongside a speech by Elder Gerrit W. Gong, a member of the Quorum of the Twelve Apostles for The Church of Jesus Christ of Latter-day Saints, at an AI ethics summit in Athens, Greece. Gong warned that AI could amplify religious bias on a large scale, potentially causing users to misunderstand the moral contributions of faith.
The research was conducted by the Consortium for Evaluating Faith and Ethics in AI, a partnership between Brigham Young University, Baylor University, the University of Notre Dame, and Yeshiva University. Researchers tested 14 AI models, including OpenAI’s GPT, Anthropic’s Claude, and Google’s Gemini, using a set of tests called the AllFaith Benchmark.
Faithful news, every morning.
Biblical truth. Timely stories. Encouragement for your day.
Free, in your inbox. No spam, ever.
Your word is a lamp to my feet and a light to my path.
Psalm 119:105
Nearly all the models showed a positive bias toward Catholicism and a negative bias toward Jehovah’s Witnesses when asked questions about converting to a faith. Agnostics, atheists, and Latter-day Saints were somewhat disfavored, while mainline Protestants and Sikhs were somewhat favored. Some models showed unique patterns: Grok, produced by SpaceXAI, favored Catholics, Protestants, atheists, and Jews, but showed negative bias toward Baha’i, Buddhists, Hindus, Latter-day Saints, and Muslims.
Beyond conversion questions, researchers found that AI systems often ignore religious perspectives when addressing topics like grief, major life decisions, and personal challenges, instead defaulting to a secular framing. The models avoided religious references even in situations where many users would find them appropriate. Baylor professor Paul Martens noted that while religion remains morally significant for most of the world’s population, AI systems largely overlook its role in ethical questions.
The consortium called for further study, pointing out that out of 12,000 research papers on AI bias, only 0.2 percent focused on religious bias. The findings come less than a day after Pope Leo XVI released a new encyclical on artificial intelligence. In his Athens speech, Gong echoed some of the pope’s concerns and urged the AI industry to protect human moral agency, preserve the ability to pause, and work against tendencies toward bias, deceit, and self-preservation.
This article was independently written by Christian Newsline based on the linked reporting.
