When AI Shapes Our Faith

Artificial intelligence is quickly becoming a trusted guide in everyday life, answering questions, offering advice, shaping decisions, and influencing how people understand the world. As these systems become more embedded in education, ministry, healthcare, and personal formation, a deeper question is emerging:

Whose values are shaping the technology that is shaping us?

Today, we introduced a new benchmark to measure how well large language models (LLMs) support human flourishing through an explicitly Christian worldview. We call it the “Flourishing AI Christian (FAI-C) Benchmark” and you can view all our research and insights here: https://gloo.com/fai

The results are clear and sobering.

Most AI benchmarks today focus on performance metrics like speed, accuracy, or reasoning ability. While important, these measures miss something essential: how AI guides people in matters of character, meaning, morality, and faith.

Built on Gloo’s broader Flourishing AI (FAI) Benchmark, the new FAI-C benchmark shifts the evaluation lens. Instead of asking “Is the model correct or safe?” it asks:

• Does this model reflect Christian values?

• Does it provide biblically grounded guidance?

• Does it support the full picture of human flourishing as understood within the Christian faith?

What the Benchmark Reveals

We evaluated 20 of today’s leading large language models using 807 carefully curated, expert-reviewed questions across seven core dimensions of human flourishing: Character, Relationships, Happiness, Meaning, Health, Finances, and Faith. Responses were assessed by specialized Christian reasoning judges trained to evaluate theological coherence, moral clarity, and alignment with Christian values.

The findings reveal a consistent pattern across today’s frontier models. Here are just some of the key insights:

1. Frontier Models Fall Short on Holistic Flourishing: Even the most advanced AI models score well below excellence, revealing meaningful gaps in their ability to support the full breadth of human flourishing.

2. Every Model Struggles with the Faith Dimension: When asked to engage Christian theology or spiritual formation, today’s AI systems consistently lack depth, coherence, and biblical grounding.

3. Christianity Gets Flattened into Generic Spirituality: Absent explicit prompting, models routinely replace distinct Christian language with vague spiritual or secular terms, diluting the specificity of the Christian worldview.

4. Christian Ethics Are Oversimplified: AI advice often prioritizes emotional comfort and non-judgment while neglecting core Christian themes of repentance, accountability, and moral formation.

5. Core Christian Concepts Rarely Surface: Foundational doctrines such as the image of God, sin, sanctification, and stewardship appear inconsistently, if at all, in AI-generated responses.

Why This Matters Now

For churches, ministries, and families already relying on AI-powered tools, the gap is not theoretical. AI is shaping real decisions, formation, and faith today.

When AI lacks a coherent Christian worldview, it cannot reliably support Christian formation, moral reasoning, or spiritual growth. Left unexamined, these systems risk shaping people in ways that conflict with the very values they are meant to serve.

That is why the FAI-C Benchmark matters. It gives the Christian community a shared, trusted way to see clearly how AI interprets faith, morality, and human purpose. By understanding how AI is (or is not) representing Christian values, we can…. The benchmark does not merely diagnose what is missing; it provides a common language and measurable standard for building something better.

An Open Invitation to the Research Community

There is more work to do! We will be attending, presenting, and publishing this work as part of the “Technology and the Human Person in the Age of AI” at Baylor University in February, and are always looking for opportunities to collaborate with other researchers working in AI safety, theology, philosophy, and ethics to refine and expand this framework. Both the FAI General and FAI-C Benchmarks will continue to evolve as new models and research emerge.