Well a polynomial $p$ of degree $n\geq1$ with rational coefficients can be thought of as a point in $\Bbb Q^{n+1}$, so say $p(x)=x^2-2\iff (1,0,-2)\in \mathbb{Q}^3$. Now if we want to count the roots would could simply consider them as points in $\Bbb Q^{n+1}\times \Bbb N_*$, where the last index will number the root, and we'll give the last number a subscript $*$, so as not to cause confusion between points. So up to some arbitrary numbering say for $p$ above we'll have two points $p_1=(1,0,-2,1_*)\iff \sqrt{2}$ and $p_2=(1,0,-2,2_*)\iff -\sqrt{2}$. Now it is also true that the cartesian product of a finite number of countable sets is countable, so each set $\Bbb Q^{n+1}\times \Bbb N_*$ is countable.
So for purely counting arguments you could consider the roots of polynomials as such a disjoint union of sets $\sqcup_{n=1}^\infty \left(\mathbb{Q}^{n+2}\times \Bbb{N}_*\right)$, and a countable union of countable sets is countable.
Of course you are over counting since the polynomials aren't all irreducible, so the same roots appear more than once, and also you have polynomials which have complex roots that are included, and also because you aren't using the entirety of the natural numbers for each degree of polynomial.
However since a subset of a countable set is countable, we're still fine as we can find an injection from our set of algebraic real numbers and the above mentioned set (the disjoint unioned one), send the algebraic number to the point associated to it's minimal polynomial, and the number assigned to that root.
Sidenote Now this is generally not the way you'd think of polynomials in algebra. Usually you'd think of the polynomial ring $\Bbb{Q}[x]$ over the field $\Bbb Q$, where the elements are written $$(1,0,2,-\frac{1}{2},0,0,\ldots)\iff 1+0\cdot x+2\cdot x^2-\frac{1}{2}\cdot x^3,$$ the dots meaning the rest are zero. But again usually the ring only contains elements whereby after some point all the entries are zero, so the same argument applies.