Class AB, it pays to experiment with NPN on top, PNP on bottom... and vice versa.
The aim is to avoid null distortion, although it may be necessary to waste some current in doing so.
When PNP is on top, NPN at bottom, it is more difficult to prevent wasted current during quiescent mode. Examine bias current, and choose resistor values so that waste is minimized.
In some cases, apply voltage from a supply rail through a resistor, to assist turn-On or turn-Off of the transistor. A potentiometer makes it easy to find a good operating point.
You'll find some designs containing diodes in strategic locations, or low-ohm resistors in emitter legs, etc. These are usually put there for a good reason even though it's not always obvous.
A surefire method is to apply a wide-swinging voltage to each transistor. See if you get expected maximum output. Reduce bias amplitudes gradually, and eventually you should see the proper range of bias voltages that produces desired performance.
Use the X/Y oscilloscope mode, to compare input signal with output waveform. You'll see a straight diagonal line when output is an undistorted copy of input. Distortion might appear in the form of clipping, non-linear (curved) plots, 'dog-legs' caused by null distortion, etc.