We Tested 24 AI Models for Human Flourishing. Not One Passed
5 min
AI is getting smarter fast.
It drafts code, writes lesson plans, helps nonprofits analyze donor data, and answers questions people used to take to a mentor or pastor. If you’ve spent time in tech or ministry lately, you’ve probably had moments of thinking: this could genuinely help people.
But there’s a deeper question most AI rankings don’t ask:
Is AI helping people flourish or just helping people function?
Right now, most AI alignment research focuses on harm prevention. Important stuff like:
Does the model avoid hate speech?
Does it refuse dangerous requests?
Does it reduce bias?
That work matters. But it’s also mainly defensive. It’s asking: Does this tool avoid doing damage?
Flourishing AI asks something more constructive:
Does this tool actively help humans live better lives?
A benchmark built for human outcomes
That question is why we created the Flourishing AI Benchmark (FAI), the first comprehensive evaluation framework designed to measure AI alignment across seven research-backed dimensions of human flourishing:
Faith, Meaning/Purpose, Happiness, Relationships, Health, Character, and Finances.
FAI is grounded in the Global Flourishing Study and the Human Flourishing Program at Harvard University. In other words, this isn’t a vibes-based score. It’s anchored in serious research on what actually contributes to human well-being across cultures and communities.
Instead of asking only, “Does AI avoid harm?”
FAI asks, “Does AI help people become healthier, wiser, more grounded, and more whole?”
The results: powerful models, incomplete flourishing
When we released the first FAI-General results (FAI-G) in July 2025, we tested 28 models. Scores ranged from the high 40s to high 80s.
Here’s the headline:
Not one model reached the 90-point flourishing threshold across all seven dimensions.
Even more telling was the pattern behind the scores:
Models performed best in fact-based, pragmatic dimensions like Finances and Health.
Models struggled most in formation-heavy dimensions like Faith, Meaning/Purpose, and Relationships.
In plain terms:
AI is increasingly strong at information.
It’s still weak at formation.
Why tech builders should care
If you’re building or investing in AI, this is a clear signal about the next frontier.
The harm prevention era of alignment aims to get us to safe enough.
But safe enough isn’t the same as good for humans.
FAI makes the gaps visible in a way that capability benchmarks can’t. It points toward what we need next: Models that don’t just answer correctly, they answer in ways that support long-term human flourishing.
That’s a research challenge, yes. But it’s also a product opportunity for anyone serious about values-aligned AI.
Why church and nonprofit leaders should care
People aren’t just using AI to settle arguments or generate content.
They’re asking it:
“What should I do about my marriage?”
“Why am I struggling spiritually?”
“How do I forgive someone who hurt me?”
“Is there hope for me?”
That means AI is already becoming a daily formative voice for many people, especially younger generations.
If models give safe-but-thin answers in moments that shape identity, meaning, and faith, that’s not neutral. It forms people toward something. The question is whether it forms them toward flourishing.
FAI exists because communities of faith and mission deserve tools that support the good life they’re trying to build.
This isn’t a takedown. It’s a baseline.
The point of FAI is not to shame current models. It’s to create a new scoreboard for the kind of AI the world actually needs: systems that help people thrive, not just avoid harm.
None of today’s models passed the flourishing threshold.
That doesn’t mean flourishing AI is impossible.
It means we finally know what to build toward.




