We show that OLS and GLS are asymptotically equivalent in the linear regression model with AR (p) disturbances and a wide range of trending regressors_ and that OLS based statistical inference is still meaningful after proper adjustment of the test statistics.