Can We Trust AI with the Truth About Faith?

Artificial intelligence is rapidly becoming the gateway to knowledge. By 2028, as many people will seek answers from AI as from Google. But when it comes to the most important questions of life - Who is Jesus? Did He rise from the dead? Is the Bible reliable? - can we trust AI to give answers that faithfully represent the Christian tradition?

A new study from The Gospel Coalition and The Keller Center for Cultural Apologetics has released the AI Christian Benchmark, the first systematic evaluation of how leading AI platforms answer theological questions.

What the Study Found

Seven top AI systems, including OpenAI’s GPT-4o, Google’s Gemini, Anthropic’s Claude, Meta’s Llama, xAI’s Grok, Perplexity, and China’s DeepSeek, were tested against the seven most Googled theological questions. Seven Christian scholars graded the results against the Nicene Creed and historic Protestant confessions.

The findings were striking:

• Two platforms (DeepSeek R1 and Perplexity) generally guided readers toward Christian faith.

• Three platforms (Grok 4, Claude 4 Sonnet, and Llama 3.7) often guided readers away from faith.

• Two platforms (Gemini 2.5 and GPT-4o) adopted an “all sides” approach, presenting Christianity as one option among many.

In terms of overall reliability, DeepSeek R1 ranked first, closely followed by Perplexity. Meanwhile, Meta’s Llama 3.7 placed last, often providing vague or skeptical responses.

Why the Results Differ

Technically, these platforms share similar architecture, training data, and computing power. So why such different answers? The study points to alignment processes, the human-in-the-loop adjustments that shape how AI responds to sensitive questions.

Alignment is designed to filter out harmful or biased content, but in practice, it often imports the values and worldview of the alignment team itself.

For example, many models default to hedging language, providing “all sides” responses that dilute Christian truth claims. Others even include nearly identical pre-written disclaimers in their answers, clear signs of human intervention.

Why It Matters for Christians

Christians cannot afford to be naïve about how AI mediates knowledge. These systems have embedded bias; they reflect cultural currents and corporate philosophies. As AI becomes the default tool for spiritual seekers, the theological orientation of its answers will profoundly shape public understanding of Christianity.

The benchmark reminds us of two key truths:

• We must be discerning users of technology. Do not assume AI is objective; give prompts context and measure answers against Scripture.

• We must engage, not retreat. If Christians do not participate in shaping how these tools handle matters of faith, others will.

As the report notes, technology itself is a form of common grace. But like every human tool, AI must be tested, refined, and redeemed for God’s glory.

At Gloo, we are contributing by continuing to develop our own Flourishing AI (FAI) Benchmark based on these same ideas. The FAI Benchmark evaluates how well AI models support holistic human flourishing — across seven dimensions such as Character, Relationships, Meaning, Health, Finances, and Faith — using over 1,200 rigorously sourced questions.

AI is already shaping spiritual conversations. The real question is whether Christians will be ready to guide, critique, and steward it toward truth?