It doesn’t take much to fit my definition of generative. An arpeggio qualifies. Really, anytime there is musical content being generated by some kind of algorithm, causing more sound to be heard than is directly being played in real-time by a person. I guess it’s distinct from a simple playback of a recording because it is generated and rule-based.
I like to use generative techniques for repetitive/rhythmic elements, and I also use generative techniques to reinforce harmonic relationships (staying in tune, chord progressions, etc). There are also times when generative techniques are handy for creating a “non-human” musical effect (creating pieces that aren’t likely to be played manually, whether due to difficulty or just strangeness).
Generative techniques are also handy for multiplying the number of “hands” I have. As a solo player, I need ways of making myself sound like more than one person.