It is the probability of observing results at least as extreme as the sample, assuming the null hypothesis is true.

A small p-value suggests the observed data is unlikely under the null, but it does not prove the alternative hypothesis or measure practical importance.