Language models do not have any concept of truth or meaning. They will state falsehoods because the words are statistically likely to appear in that order. You should always verify the output unless the only thing you need it to be is grammatically comprehensible.