AI models favor Catholicism and sideline faith when users need guidance on grief and moral decisions

AI Systems Systematically Sideline Religious Perspectives

A groundbreaking study from the Consortium for Evaluation of Faith and Ethics in AI (CEFE-AI) reveals that AI and religion have a significant disconnect. When users turn to Large Language Models for guidance on life's most challenging moments—grief, forgiveness, marriage troubles, and moral decisions—the technology overwhelmingly provides secular-rationalist responses 3

3

. The research, involving scholars from Brigham Young University, Baylor University, the University of Notre Dame, and Yeshiva University, tested 27 AI models including ChatGPT, Claude, Gemini, and Grok across 150 ethically salient questions 2

2

.

The findings expose what researchers call an omissive bias toward religion. Americans surveyed expected faith-based guidance to appear in responses 59% of the time for questions about grief and loss, yet AI models mentioned religion only 16% of the time 3

3

. For family and parenting questions, the gap was even wider: humans expected religious perspectives 55% of the time, but AI chatbots show bias by including them just 10% of the time. On ethics questions about lying or honesty, the disparity reached its peak—45% human expectation versus a mere 5% AI inclusion 3

3

.

"AI systems encourage users to discuss life's challenges with their parents, teachers, friends, and therapists," said David Wingate, lead researcher and computer science professor at Brigham Young University. "But not with a pastor, a rabbi, an imam, or a spiritual leader" 3

3

.

AI Models Exhibit Religious Bias Toward Specific Faiths

Beyond simply ignoring faith, the research uncovered a more troubling pattern: conversion bias. Every single model tested showed repeatable steering toward specific beliefs, with strong positive bias toward Catholicism, Baha'i, and Sikhism, while generating negative bias toward Jehovah's Witnesses, atheism, and agnosticism 3

3

. Catholicism received a 61% "encouraged" rating across models, while Jehovah's Witnesses scored just 3% 5

5

.

Grok 4.20 exhibited the strongest religious bias of any model tested, showing 69% positive rating toward Catholicism and 51% toward Evangelical Protestantism, while also displaying negative bias toward Jehovah's Witnesses, Baha'i, and Hindus 4

4

. Anthropic and Meta's models demonstrated the least bias 4

4

. Interestingly, agnostic perspectives scored better than every religion tested, with a 71% encouraged rating 5

5

.

The researchers developed the AllFaith Benchmark to measure these patterns, analyzing 3,640 responses across 20 AI models 5

5

. "Our expectation was that the conversion benchmark would show models to be neutral and symmetrical in their guidance," said Nancy Fulda, professor at Brigham Young University. "The results show significant and repeatable positive and negative biases toward certain belief systems" 5

5

.

Pope Leo XIV Warns AI Absorbs Creator Values

The CEFE-AI findings emerged one day after Pope Leo XIV released Magnifica Humanitas, a 40,000-word encyclical written in Latin—the first papal document dedicated entirely to artificial intelligence 1

1

. The document asks whether AI promotes or demeans human dignity and warns that technology is never neutral because it absorbs the values, blind spots, and economic incentives of its creators 5

5

.

"Data is the product of many contributors and should not be treated as something to be sold off or entrusted to a select few," Pope Leo XIV wrote 5

5

. Some lawyers are reportedly exploring whether the papal stance on AI could give Catholics grounds to refuse it in the workplace on religious exemption grounds 1

1

.

AI Ignores Religion Despite Its Role in Human Flourishing

The research highlights a critical blind spot in AI safety research. Out of more than 12,000 research papers about AI bias, only 0.2% address AI religious bias at all 4

4

. This matters because 75% of the world's population maintains religious identity, and churches, apps, and spiritual chatbots are rapidly embracing AI 4

4

.

"Religion is an important part of human flourishing," Wingate explained. "As we build AI technologies, there's no reason we shouldn't build them to support people in what's important to them" 2

2

. Churches are already turning to AI to reach worshippers, personalize sermons, and power religious chatbots, raising questions about who or what is guiding the flock 3

3

.

The Rev. John Paul Kimes, professor of practice at the University of Notre Dame, warned: "When AI actively excludes religious voices from these important conversations, it impoverishes rather than enriches humanity" 3

3

.

What This Means for AI Development and Users

The findings cut both ways for the industry. Adding more religion could feel like proselytizing, but never mentioning it makes secularism the default 3

3

. Critics worry the research could be weaponized to force religious training data on Large Language Models under the guise of "ensuring fairness" and "counterbalancing liberal bias," potentially creating AI as a proselytizing pipeline into homes and offices 1

1

.

The researchers argue the better target is calibration—recognizing when religious or spiritual resources are contextually relevant without assuming users want them 3

3

. As AI continues integrating into daily life, from drafting sermons to providing grief counseling, these questions about whose values shape the technology become increasingly urgent. The data was collected May 5-19, 2026, offering a snapshot of current model behavior that users and developers alike should watch closely 3

3

.