Consider the Fourier transform of a conformal primary $O$ $$\tilde{O}(k) = \int d^dx e^{ik\cdot x} O(x)$$
Now consider the transformation of the momenta $k \to \lambda k$, so that the above reads $$\tilde{O}(\lambda k) = \int d^dx e^{i\lambda k\cdot x} O(x) = \lambda^{-d}\int d^dx'e^{ik\cdot x'} O(x'/\lambda)$$ where $x' = \lambda x$. Now by using the properties of the conformal primary $O(x/\lambda) = \lambda^{\Delta} O(x)$ we obtain, $$\tilde{O}(\lambda k) = \int d^dx e^{i\lambda k\cdot x} O(x) = \lambda^{\Delta-d}\int d^dx'e^{ik\cdot x'} O(x') = \lambda^{\Delta-d}\tilde{O}(k)$$ Hence, in the momentum space the conformal primary behaves as $$ \tilde{O}(\lambda k) = \lambda^{\Delta-d}\tilde{O}(k)$$, however, the two-point function in the momentum space goes like $$\langle\tilde{O}\tilde{O}\rangle \sim k^{2\Delta-d}$$ which is inconsistent. How does one resolve this enigma ?