I am also pretty new to simulation, but I think I can answer your question. The thing is that in framework of Ardupilot Gazebo is usually used in combination with SITL, i. e. you can use SITL without Gazebo but you will hardly be able to use Gazebo without SITL.
You should treat SITL more like an engine to model kinematics and dynamics the vehicle and environment, and to engage with the firmware, while Gazebo is responsible for 3D modeling and rich visualization. Hence, if you want to explore some basic stuff about your firmware without need for 3D interaction with environmental object - you can make it just with SITL. If you want to check some 3D environmental perception and interaction characteristics for your firmware - you should anyway start with SITL and then, when you feel that you are already comfortable with it, move to adding Gazebo to the simulation.
I am not sure, that my answer is 100% precise, but I thing that it is very close to the truth.