Every AI gives answers. We audit them.

Paste an answer from ChatGPT, Claude, Gemini — any assistant. Second Opinion reads it the way a calibration analyst would: which claims are flatly asserted, which are actually checkable, which could be stale — and the fastest concrete way to verify each one before you act on it.

Honest banner: this audit is itself produced by a language model — a structured, deliberately skeptical read, not ground truth. It judges how the answer frames certainty; it usually cannot know whether claims are true. Nothing you paste is stored with your identity; the free lane is capacity-limited and may ask you to retry at peak. For Brier-tracked answers built this way from the start, see the AEQUARA tools.

Related: translate the hedge-words it used with Words → Numbers · lock your own call with VERD Receipts · measure yourself on the Scorecard.