Spectral decimation is an inductive process where the eigenvalues of a natural Laplacian on "nice" fractals is computed inductively. The idea is that by using a sequence of finite graphs to approximate the graph and graph Laplacians on these graphs one can inductively determine the eigenvalues of the Laplacian on the n+1'st graph from those of the n'th graph. Ultimately the spectrum of the limit Laplacian is the closure of the scaling limit of the spectra on the finite graphs. The original reference for this is [here][1]. (Apologies to those who do not have access to MathSciNet). 

I am not sure which side of the induction versus iteration divide this lands for you. Certainly much of the emphasis on self-similar fractals is an attempt to be able to call upon iterative methods at the very least and in many cases these proofs do become inductive. There is another proof that I am thinking of that occurs in proving Harnack inequalities for harmonic functions on the Sierpinski carpet, but that is for another answer.

The local definition of "nice" for this answer is either post-critically finite fractals or finitely ramified with a doubly transitive symmetry group fractals. 


  [1]: http://www.ams.org/mathscinet/search/publdoc.html?arg3=&co4=AND&co5=AND&co6=AND&co7=AND&dr=all&pg4=AUCN&pg5=AUCN&pg6=PC&pg7=ALLF&pg8=ET&review_format=html&s4=fukushima&s5=shima&s6=&s7=&s8=All&vfpref=html&yearRangeFirst=&yearRangeSecond=&yrop=eq&r=3&mx-pid=1245223