What is the best way to approach testing for Ask an AI Question?

I am trying to improve my approach to Ask an AI Question and could use some practical advice.

I am particularly interested in testing and automation. I would like to compare approaches that people have actually used rather than just theoretical suggestions.

Which approach would you recommend and why?