A theology quiz with teeth

Most people have never been asked what they actually believe.

Which Pew Do You Skew is a tradition classifier that maps your theological instincts to 14 historic Christian traditions — and catches you when you contradict yourself.

Take the Quiz →
The Backstory
· · ·

The problem with theology quizzes

Most denominational quizzes are personality tests wearing a clerical collar. Ten vague questions, a single score axis, and a result that tells you you're "62% Presbyterian" without explaining what that commits you to or where your answers don't actually fit together.

The real question isn't "which denomination matches your vibes." It's: do you actually hold a coherent theological position, and if so, which historic tradition does it most closely resemble? Those are two very different questions, and most quizzes don't bother with the first one at all.

We wanted a quiz that took the theology seriously — one that could distinguish between a Reformed Baptist and a Presbyterian on the questions that actually matter, catch a Catholic who accidentally affirmed sola scriptura, and tell someone honestly when their answers don't form a coherent worldview.

· · ·

Built in a single session

In January 2026, I wrote a spec for a two-layer quiz — orthodoxy gate first, then tradition classifier, with a contradiction engine that catches incoherent answers. I knew the architecture I wanted because I'd spent years watching theology quizzes get it wrong: they sort you into a tradition without first checking whether you affirm the creeds, which means someone with Arian Christology can land as "Methodist" because they vibed through the soteriology questions.

The spec went to Claude to build. By that evening, the first working prototype existed as a single HTML file — scoring matrix, gate logic, contradiction checks, and all 14 traditions wired up. The theology was mine. The code was the AI's. That division of labor held for the entire project.

"Something I could practically host online in basically one file."

— The entire technical spec
· · ·

Two layers, one engine

The quiz runs in two passes. The first pass is a gate. The second pass is a classifier. They serve different purposes and the order matters.

Layer 1 — The Gate

Six questions testing creedal orthodoxy: Trinity, Incarnation, canon of Scripture. If you affirm Nicaea and Chalcedon, you pass. If not, you get routed to one of four non-creedal outcomes — Non-Trinitarian, Modalist, Arian, or Restorationist — and the tradition classifier never runs. This prevents the awkward result where someone with Arian Christology gets sorted into "Methodist" because they happened to answer the soteriology questions a certain way.

Layer 2 — The Classifier

27 questions across authority, sacraments, soteriology, ecclesiology, eschatology, and ethics. Each answer carries weighted scores across all 14 traditions. Your result is the tradition with the highest cumulative score, adjusted for coherence and confidence.

Layer 3 — The Contradiction Engine

19 defined contradiction checks run against your completed answers. Some are hard (believing in pre-tribulation rapture and postmillennialism simultaneously), some are soft tensions (strong third-use-of-the-law conviction with a Lutheran match). The quiz doesn't just sort you — it catches you.

· · ·

The quiz that argues back

The contradiction engine is the thing that separates this from a Buzzfeed quiz. When your answers don't fit together logically, the quiz tells you — and explains why.

Contradiction Detected

You affirmed believer's baptism only and that children of believers are members of the visible church.

If children are covenant members, most traditions that hold that view also baptize them as a sign of that membership. Holding both positions creates a tension you'd need to resolve — either children aren't really "in" the covenant community, or the sign should follow the status.

This matters because most people have never been forced to hold their positions next to each other and check whether they're actually compatible. You can believe in congregational autonomy and papal infallibility in the same quiz — but the quiz won't pretend you didn't.

When the contradictions pile up past a threshold, the quiz stops claiming a tradition for you entirely. Instead of a false match, you get an honest "Eclectic Profile" — your answers are real, but they don't belong to any one historic tradition. That's not a failure state. It's a starting point.

· · ·

33 questions, 14 traditions

33
Questions
14
Traditions
19
Contradiction
Checks
6
Gate
Questions
Roman Catholic Eastern Orthodox Lutheran Anglican Reformed / Presbyterian 1689 Reformed Baptist General / Evangelical Baptist Methodist / Wesleyan Non-denominational Pentecostal / Charismatic Dispensational Evangelical Anabaptist / Mennonite Quaker / Friends Seventh-day Adventist
· · ·

The hardest answer is "I don't know"

Early testers kept vibe-checking their way through questions they'd never actually considered. "How does God act in salvation?" got answered with gut instinct instead of conviction, and the results reflected vibes at 9:42pm rather than a worldview.

The fix: "I'm not sure" is a first-class answer on every classifier question. It doesn't dodge — it degrades your confidence score honestly. If you mark too many pivotal questions as undecided, the quiz tells you which ones matter most for determining your tradition and invites you to think about them. Not as a penalty. As a map of what you haven't decided yet.

The quiz surfaces the questions you've never been asked — and tells you which ones actually matter for where you'd land.

· · ·

How it got here

January 2026
The spec
Two-layer architecture: orthodoxy gate, then tradition classifier, with a contradiction engine. I wrote the theological framework and scoring logic; AI coded the implementation. Built as a single HTML file in one evening. 28 questions, 13 traditions, 10 contradiction checks.
The next few days
Stress-testing the question set
I ran the question list through multiple AI tools looking for coverage gaps. The Anabaptist/Mennonite tradition was a real hole — the "sword" question cleanly separates them from Quakers. Added the millennium question (which I'd originally scoped but cut for v1) and Third Use of the Law as a sharper Lutheran discriminator. Question count climbed to 33.
Iteration
The "I don't know" problem
Multiple versions of the unsure-answer handling. Settled on dynamic injection of "I'm not sure" to every classifier question, honest confidence degradation, and an undecided positions summary.
February–March 2026
The audit gauntlet
I ran the scoring matrix and contradiction logic through four separate AI tools as adversarial QA — checking for miscalibrated weights, missed edge cases, and data entry errors across hundreds of score assignments. Three of the four flagged the Reformed/Lutheran fork penalty as too aggressive. I kept it, because the sacramental watershed really is that sharp and they were wrong.
Launch
"Theological Tradition Finder" becomes "Which Pew Do You Skew?"
The original working title sounded like a DoD briefing. The final name landed on its own. A single static file, self-hosted, free forever. Because a theology quiz that charges admission has already lost the plot.
· · ·

This is a map, not a membership card

The quiz doesn't tell you what to believe. It tells you what you already believe — and shows you the tradition that thought it first.

Most people walking through a church door have a theology. They just haven't named it. They know what they think about baptism but couldn't tell you which confession agrees with them. They have instincts about authority, about salvation, about what the Lord's Supper means — but those instincts live as feelings, not as positions mapped to a tradition with 500 years of thought behind it.

Which Pew Do You Skew takes the feelings and gives them addresses. It connects your instincts to the people who've already thought them through, written them down, and argued about them for centuries. That's not sorting. That's introducing you to your own convictions.

The goal was never to put people in boxes. It was to show them the box was already there — they just hadn't read the label.

· · ·

From the same pew

Know where you land? Now see if you can predict the sermon. Pulpit Bingo is a sermon prediction game built on the Revised Common Lectionary. Same builder, different game. Read its backstory.