Because graphical programming are just textual programming languages which are just overly complex (xml) and coming with attached visualization engine for its components. They are used only for limited domains such as composing some data flows or video node editors, etc. because visualization of anything more sophisticated would make one's head explode. You could just as well visualize components of AST of Python for very little benefit.

The only upside of graphical languages is that because of their limited capabilities, they focus on only one abstraction layer and that allows you to visualize that layer quite cleanly.

You can't program anything more sophisticated or step outside of the only abstracrion layer you are allowed to program in. Eg. you can't write your custom for loops by dragging and dropping boxes and you also cannot orchestrate dags or box diagrams using dragging other boxes around.