BALTIMORE, July 28, 2026—The Institute for Islamic, Christian, and Jewish Studies (ICJS) is developing a new AI benchmark to measure how different large language models, or LLMs, handle interreligious questions.
As AI increasingly becomes a place where people seek information and guidance about diverse religions, ICJS is asking a timely question: How are AI platforms currently responding to interfaith and interreligious inquiries? A benchmark is a standardized way to measure and evaluate the platforms’ responses to a diverse set of questions about Islam, Christianity, and Judaism.
The ICJS Interfaith Benchmark 1.0 is expected to be an important tool to assess the efficacy of different chatbots in improving interfaith understanding for users seeking religious information–ranging from the theological to the practical.
“People who are curious about the beliefs, practices, and ethics of diverse religious communities will bring their interreligious questions to AI,” said Heather Miller Rubens, Ph.D, ICJS executive director and the project’s lead investigator. “This benchmark is an opportunity to think carefully about how AI is answering those questions and how these emerging technologies might be better designed to support interreligious understanding.”
This summer, Rubens is being assisted by Jacob Green, a Ph.D. student in media studies at the University of Virginia. His internship at ICJS is funded by Templeton Religion Trust through a new American Council of Learned Societies initiative supporting graduate internships with organizations by doing practical research around the “Spiritual Infrastructure of the Future.”
A benchmark is essentially a standard test researchers use to evaluate LLMs, including ChatGPT, Claude, and Google Gemini, by measuring their responses on a designated task. The ICJS Interfaith Benchmark 1.0 will pose identical questions to each LLM, examine the answers side-by-side, and score the responses using a weighted rubric that evaluates accuracy, tone, bias, and ability to support constructive dialogue. The benchmark results can be used by both AI developers to improve LLM performance and by the general public to understand how LLMs respond to interreligious and interfaith questions.
Aligned with the mission and values of ICJS, the Interfaith Benchmark 1.0 will measure any disparities in how different religions are treated, evaluate which questions are the most susceptible to religious bias and bigotry, and to encourage AI developers to see interfaith understanding and engagement as a worthy societal goal for this widely-used technology.
About ICJS
To disarm religious bias and bigotry, the Institute for Islamic, Christian, and Jewish Studies (ICJS) builds learning communities where religious difference becomes a powerful force for good. ICJS envisions an interreligious society in which dialogue replaces division, friendship overcomes fear, and education eradicates ignorance. Through educational programming, public-facing scholarship, and relationship-centered fellowships and workshops, ICJS models a new conversation in the public square that affirms religious diversity in the United States. ICJS is an independent 501(c)3 nonprofit organization. More information is at www.icjs.org.