It requires the population to be normally distributed.

It is a non-parametric method that relies on the exchangeability of observations under the null hypothesis.

It is less robust than a $t$-test when outliers are present.

It calculates p-values using the chi-squared distribution.