Geometrically the issue is that analytic functions in a simply connected Domain are orientation preserving and conjugation is orientation reversing. So if $f$ is non-constant, e.g. where $f(0)=w$ and $\gamma=r\cdot \exp\big(2\pi i\cdot t\big)$ for $t\in [0,1]$, $r\gt 0$ small enough, the Argument Principle says
$0 \lt n\big(f\circ \gamma,w\big)=\sum_{z, F(z)=w}v_F(z)\cdot n\big(\overline f\circ \gamma, z\big)\leq 0$
as the valency $v_F(z)\in \mathbb N$ and $n\big(\overline f\circ \gamma, z\big)\leq 0$ for any $z\in \mathbb C-[\bar f\circ \gamma]$
(i.e. all $z$ in the complement of the image of $\overline f\circ \gamma$)
and this is impossible.
proof of orientation reversal for conjugation
let $f$ be any analytic function in a simply connected Domain, $\gamma$ a closed curve and the winding number $n\big(f\circ \gamma, w\big)$ be well defined. Define $g(z):=f(z)-w$. Then
$0=n\Big(\big(g(z)\cdot \overline g(z)\big)\circ \gamma,0\Big)=n\Big(g\circ \gamma,0\Big)+n\Big(\overline g\circ \gamma,0\Big)=n\Big(f\circ \gamma,w\Big)+n\Big(\overline f\circ \gamma,\overline w\Big)$
the Left Hand Side follows because $g(z)\cdot \overline g(z)\in \mathbb R_{\gt 0}$
$\implies-n\Big(\overline f\circ \gamma,\overline w\Big)= n\Big(f\circ \gamma,w\Big)\geq 0 \text{ (Argument Principle)}$
and the LHS holds for arbitary $\overline w$ in the complement of the image of $\overline f\circ \gamma$