Suppose $G$ is a group and $H < G$ and $A_H=\{g \mid \bar{g} \subset H \}$ where $\bar{g}$ is the conjugacy class of $g$. I want to prove $A_H\lhd G$.
First approach:
Define $X$ as the set of left cosets of $H$, then define a homomorphism $ \varphi(g) = f_g$ from $G$ to $S_X$ where $f_g(aH)=gaH$.
$g_1=g_2$ implies $f_{g_1}=f_{g_2}$, so $\varphi$ is well defined. Also $\varphi(g_1 g_2)=f_{g_1 g_2}=f_{g_1}(f_{g_2})=\varphi(g_1) \circ \varphi(g_2)$ shows $\varphi$ is a homomorphism.
Now, $\varphi(g)=e_{S_X}$ implies $f_g(aH)=gaH=aH$ for all $a \in G$. Equivalently, it implies $a^{-1}ga \in H$ for all $a \in G$. Therefore, $\ker(\varphi)=A_H$, and so $A_H \lhd G$.
Now, I want to directly prove, first, $A_H$ is a subgroup of $G$, then it is normal in $G$.
I cannot even prove $A_H$ is a subgroup because I cannot prove $A_H$ is closed under inverse.
Can anyone prove $A_H \lhd G$ directly?
I will use the one-step subgroup test.
By definition, $A_H\subseteq G$.
For all $g\in G$, we have $geg^{-1}=e\in H$, so that $\overline{e}=\{e\}\subseteq H$. Thus $e\in A_H$. Thus $A_H\neq\varnothing$.
Let $a,b\in A_H$. Then $\overline{a}, \overline{b}\subseteq H$. Consider $\overline{ab^{-1}}$: it is the set of all $k\in G$ such that $k=g(ab^{-1})g^{-1}$ for some $g\in G$; but then
$$k=(gag^{-1})(gb^{-1}g^{-1})\in \overline{a}(\overline{b^{-1}}),$$ where $\overline{b^{-1}}$ is a subset or equal to $H$, since $H$ is closed under inverses and $\overline{b}\subseteq H$ (specifically: $gb^{-1}g^{-1}=(gbg^{-1})^{-1}\in H$). But the set product $AB$ of any $A,B\subseteq H$ is in $H$ as $H$ is itself a group. Therefore, $\overline{ab^{-1}}\subseteq \overline{a}(\overline{b^{-1}})^{-1}\subseteq H$. Hence $ab^{-1}\in A_H$.
Hence $A_H\le G$.
For normality, let $g\in G$ and $a\in A_H$. Then $\overline{a}\subseteq H$ and so
$$\begin{align} \overline{gag^{-1}}&=\{d(gag^{-1})d^{-1}\mid d\in G\}\\ &=\{(dg)a(dg)^{-1}\mid d\in G\} \end{align}$$
is simply $\overline{a}$ because $\rho_g:G\to G$, given by $q\mapsto qg$, is a bijection. Therefore, $\overline{gag^{-1}}\subseteq H$, so that $gag^{-1}\in A_H$.
Hence $A_H\unlhd G$.