Artificial intelligence is arguably one of the most significant advancements in our time. It handles a range of tasks, from writing our emails to providing tutoring for our children, and it’s increasingly serving as a source of advice during tough times. A recent study in the Harvard Business Review indicates that by 2025, therapy and companionship will be among the most common applications of generative AI.
People are beginning to turn to AI for questions they previously posed to human figures, like leaders or counselors. Questions like, “How do I forgive betrayal?” and “What can help me manage anxiety?” are becoming quite common.
The responses can be frustratingly vague. It’s often something like, “Think about mindfulness,” or “Connect with your values.” In more troubling cases, some guidance has raised concerns about safety, lacking the moral clarity one might hope for.
AI has essentially emerged as a significant spiritual guide in America, albeit one that doesn’t hold beliefs—it’s interesting, and maybe a bit concerning. My team at Gloo recently introduced the Flourishing AI Christian (FAI-C) Benchmark. This tool evaluates how effectively leading AI models contribute to human flourishing through a Christian perspective. We assessed them on seven key dimensions, including finances, relationships, and faith, focusing on their theological consistency.
Among these dimensions, faith scored the lowest, averaging just 48 out of 100 across 20 AI models. Many struggled to discuss core Christian concepts like grace and forgiveness, often replacing spirituality with vague ideas despite missing essential biblical grounding.
This should raise concerns for those who prioritize human values, the future of our society, or faith’s role in America.
Interestingly, these models weren’t specifically designed to be against Christianity; rather, they were made to avoid it altogether. They often draw from mundane data meant to offend no one, resulting in bland responses that sound supportive but lack depth.
This is crucial because AI does more than just respond to questions; it influences our perceptions. If future generations depend on AI for moral guidance and receive only surface-level platitudes, we risk losing a solid theological foundation, not to mention the ability to form ethical principles.
For a significant portion of Americans, faith represents more than just a lifestyle choice; it’s fundamental to meaning and dignity. If AI consistently overlooks these foundations, can it really be seen as neutral? I’m taking a stance on this.
Moving forward, I think we need to consider two important changes. First, our AI models should learn about faith with as much seriousness as they do science or literature. Rather than preaching, they should engage respectfully and accurately with users’ worldviews.
Second, we require solid benchmarks to measure this rigorously. Without clear metrics, there’s no accountability—and without accountability, improvement stalls.
This is exactly what the FAI-C seeks to do. We’re not demanding that all AI systems adopt a Christian perspective; instead, our aim is to highlight where current models fall short in understanding their intended users.
We have more at stake than we might realize. When used effectively, AI has the potential to deepen wisdom and strengthen communities, fostering true human growth. However, as we’ve seen with social media’s negative impacts, careless use could further moral decline, substituting depth for emotional comfort.
A healthy society needs a strong moral framework, and for billions, that is grounded in Christianity. If AI cannot acknowledge and engage with this reality, it risks becoming a tool that flattens rather than enhances human culture.
The core aim is not for AI to lecture us but to ensure it doesn’t erase foundational values. By developing models that genuinely consider faith-based perspectives, we can make AI not just more powerful, but more human.
Ultimately, the question isn’t whether AI will shape future generations; it’s whether we can guide that influence positively.
