LANDMARK · 5 MIN · ORIENT
Fluent is not true
After this landmark, you can catch yourself trusting an answer because of how it sounds, and separate fluency from evidence.
Language models optimise for text that reads well, not text that is true. So a wrong answer arrives in the same confident, polished prose as a right one — with invented citations, realistic names, and plausible numbers. Fluency is a property of the writing, not a signal of accuracy. When being wrong would matter, the sound of the answer tells you nothing; the source does.
This is the canonical concept. It stays the same across learner lenses so personalization never changes the underlying facts.
◌ EVERYDAY LENS
What this looks like for you
If an answer would change a real decision — health, money, a claim you’ll repeat — don’t let its confidence settle it. Ask for a source you can open, or check elsewhere.
MAKE A DECISION
An AI gives you a clear, well-organised answer with three cited studies. What’s the right level of trust?
CARRY THISPick one recent AI answer you acted on. Would it survive you opening its sources? Do it once.