To make this more precise: fix a ground field $K$ and a polynomial $f(x) = a_0 + a_1x + ... + a_n x^n$ with (perhaps unnecessary) $a_n \neq 0$. Then for each possible Galois group $H$ (which one can see are the subgroups of $S_n$) is there a first-order criterion $P_H$ such that $P_H(a_0,...,a_n)$ holds if and only if the splitting field of $f$ has a Galois group isomorphic to $H$?
Or perhaps just one direction - is there a statement $P_H'$ which, when satisfied, implies the Galois group of the splitting field is isomorphic to $H$?
I think this is possible but I do not have a very rigorous argument or sketch in mind. Very very roughly, it seems like it boils down to something like ``label the roots of $f(x)$ as $\alpha_1,...,\alpha_n$ and then one can test whether a permutation of them is an isomorphism of $K[x]/f(x)$ in a first-order way'' but I'm struggling to make that more precise.
Yes. Fix $N\in\mathbb{N}$ and a subgroup $G\subseteq S_n$. Then you can write down a first-order formula which says:
Taken together, this says that $K^N$ is a splitting field of $f$ and its automorphism group is $G$. Since the splitting field of $f$ has degree at most $n!$, we can take a disjunction of these formulas over all $N\leq n!$ and all $G$ which are isomorphic to $H$ to get a formula which says the Galois group of $f$ is isomorphic to $H$.