First of all, the Riesz potential operator is not even well-defined on the Hölder class $C^\alpha$. For example, it diverges when applied to non-zero constant functions. You need to restrict to a narrower class of functions.
Second, any statement of this form requires a particular definition of $C^\alpha$ when $\alpha$ is an integer: in this case $C^\alpha$ should be defined in terms of second order differences, and it is a proper subset of the space traditionally denoted by $C^{\alpha-1,1}$.
Theorem 4 in Section V.4.4 of the classical Stein's book Singular Integrals and Differentiability Properties of Functions asserts that the Bessel potential operator $J_\alpha$ is an isomorphism between $C^\beta$ and $C^{\beta + \alpha}$. The operator $J_\alpha$ is a Fourier multiplier with symbol $(1 + |\xi|^2)^{-\alpha/2}$, instead of the symbol $|\xi|^{-\alpha}$ of the Riesz transform $I_\alpha$.
If we only consider, say, compactly supported $f$, then it is easy to switch from Bessel to Riesz potential operator: simply apply Lemma 2 from Section V.3.2 of Stein's book, which asserts that the operator $J_\alpha^{-1} I_\alpha$ is a convolution with a finite measure, and write $I_\alpha f = J_\alpha ((J_\alpha^{-1} I_\alpha)(f))$.
You will find a lot of similar results also in Samko's book Hypersingular Integrals and Their Applications.