So there's an old theorem of Lang and Weil showing that the Riemann hypothesis for curves over finite fields implies a kind of quasi-riemann hypothesis for surfaces over finite fields.

I am wondering: is it possible to go in the other direction? Namely, if we knew the Riemann hypothesis for surfaces which are *not* products of curves (so no trivial argument with Kunneth applies), would it be possible to deduce the RH for curves?

Since this is such an odd question, let me lay my cards on the table. I am looking for possible motivation for a number field analogue; namely, if the Riemann hypothesis for $GL_2$ L-functions implies anything about the Riemann hypothesis for $GL_1$.

**Edit**: If $X\to C$ is an elliptic surface which is not split, there is an isomorphism $H^1(X)=H^1(C)$, which answers my original question in a lot of cases. So let me be even more persnickity and refine my question to: if we knew RH for Frobenius acting on $H^2$ of non-split surfaces over finite fields, could we show this implies RH for curves?