Mass scaling is used in explicit analysis. The size of the smalles time increment is determined by the time a sound wave needs to traverse the smallest element in your structure. Since the speed of sound depends on density, artificially increasing the density of your material leads to larger timesteps and can thus improve efficiency.

However, this should only be done if your problem is not sensitive against the effect of elastic waves (one standard example is rolling, where explicit methods are used t overcome convergence problems, but elastic waves only play a minor role.)

Automatic mass scaling means that Ansys tries to determine an appropriate mass scale factor (the handbook should explain how this is done - I don't know Ansys so I can't explain details). I would nevertheless recommend to compare results with and without mass scaling at least for a simple test case to make sure it will not affect your solution.