In whatever text/review I happen to come across (like for example From Noether’s Theorem to Bremsstrahlung: A pedagogical introduction to Large gauge transformations and Classical soft theorems, written by Noah Miller), the explanation for the Noether's theorem(s) starts in the following way:
If you have a symmetry transformation $\delta\phi_i=\epsilon \delta_{\text{sym}}\phi_i$ where $\epsilon$ is an infinitesimal constant, then the action will change by the total derivative of some vector $J^{\mu}$.
My question is, why does the fact that a transformation is a symmetry transformation imply that the action will change by a total derivative of some vector $J^{\mu}$? I know that this is the definition of "the symmetry transformation", but what is the motivation behind this definition? Does the motivation have to do with Taylor expanding the action?
I know I have asked this earlier, but I didn't get an answer.
P.S.: I have this question because I was under the impression that a symmetry is a transformation that does not alter the classical equations of motion. But apparently, this is not the case.