How do I know if an AI answer is correct?
You cannot tell whether an AI answer is correct by reading it, because a correct answer and a confident wrong answer look identical. To actually know, make the answer prove itself: require every important claim to name a source you can open, ask the same question of another model or two and see where they disagree, and press on the disagreements until they resolve on evidence rather than on which answer sounded better. What survives that is trustworthy. What cannot name a source or falls apart under a second model's scrutiny was never safe to act on.
The reading trap
Your instinct is to judge an answer by how well it reads. That instinct is exactly backwards with AI. Fluency, structure, and confidence are the things these systems are best at, whether or not the underlying claim is true. The best-written answer is often the most dangerous one, because it earns trust it has not verified.
The three checks that actually tell you
- Can it name a source you can open? Not "studies show." An actual, checkable source. Claims that cannot name one get demoted, no matter how confident they sound.
- Does a second model agree for the same reasons? Ask another model independently. If they agree, check whether they agree on the evidence or just on the conclusion. If they disagree, you have found exactly the spot that needs a human.
- Does it survive an attempt to kill it? Ask the model, or another model, what evidence would make the claim wrong. A real claim can name its own undoing. A hallucination just insists, louder.
Do not stop at agreement
The common shortcut is to ask a couple of models and trust whatever they agree on. That is better than one model, and it is still not proof. Models trained on similar data share blind spots and can be wrong together. Agreement is a starting point, not a verdict. See if two AI models agree, does that mean the answer is right.
The full method for knowing, not guessing, is the Council Method, taught in Let the AI Be Smart.
Go deeper: this site's hub page on the Council Method is the full definition. Related questions: How do I stop AI from hallucinating?, If two AI models agree, does that mean the answer is right?, How do I fact-check an AI answer without being an expert?.