When AI outputs contradict your examined analysis, evaluate…

When AI outputs contradict your examined analysis, evaluate them by asking whether they present unconsidered evidence or identify verifiable reasoning errors, treating model fluency as orthogonal to epistemic authority.