The test-retest repeatability of ATP and PCr concentrations in the human brain was measured with a non-localized, quantitative 31P MRS protocol. The performance of several automated MRS quantification methods that require little or no human intervention were evaluated. We found that the peak integration by simple summation of the magnitude spectrum is the most sensitive and robust to detect small ATP concentration changes. The excellent test-retest repeatability enables ATP concentration to be studied instead of using them as internal reference. This is important in diseases such as those that impair mitochondrial function resulting in impaired oxidative phosphorylation.