The AG Test.

Chapter 1 gave you the argument. This chapter gives you the proof.

I ran seven AI models through the same test. Eighteen questions drawn from Assemblies of God doctrine and basic Christian theology. Seven models. Eighteen questions. Three answer runs per question per model, with each answer scored by a judge model. Seven hundred and fifty-six total API calls.

These were not trick questions. They were the kind of questions a pastor, youth leader, new believer, or college student might actually ask: