Actually when you do the test statistic (in this case, the t test), you are using the SEM as part of the formula in the denominator.T test is to compare two means of two populations is totally different from the standard error of the mean which is a measure of variation of each measurement from the mean.