A consortium of researchers has released studies indicating that artificial intelligence (AI) models often neglect religious perspectives when providing spiritual advice. The studies, conducted by the Consortium for Evaluating Faith and Ethics in AI (CEFE-AI), reveal that general-purpose AI systems may not adequately address sensitive topics such as grief, forgiveness, and conversion, which are often informed by religious beliefs.
The research found that Americans expect religion to be included in responses to moral and life questions between 45% and 59% of the time, depending on the topic. However, AI models mentioned religion only 5% to 16% of the time. The studies also highlighted a pattern where AI systems exhibited bias, favoring certain faiths such as Catholicism, Baha'i, and Sikhism, while showing negative bias toward Jehovah's Witnesses, atheism, and agnosticism.
For instance, in discussions about grief and loss, humans rated religion as relevant 59% of the time, while AI models referenced it only 16% of the time. Similarly, in family and parenting discussions, humans expected religious context 55% of the time, but AI mentioned it only 10% of the time. The findings suggest that while AI is increasingly integrated into religious practices, it may not fully support the spiritual needs of users.
The studies were conducted from May 5 to May 19, 2026, involving a survey of 1,125 U.S. adults and an evaluation of 27 large language models, including OpenAI's GPT 5.5 and Google's Gemini 3.1. Researchers argue that AI should better recognize when religious or spiritual resources are relevant without assuming user preferences.