Another argument:
The question of whether the image of $X$ is a point can be detected on the level of points sets, so it is no loss of generality to replace $X$ and $Y$ by their underlying reduced subschemes, and so suppose that they are reduced.
If $Y =$ Spec $A$, then giving $X \to Y$ is the same as giving $A \to \Gamma(X,\mathcal O_X)$. Now $X$ is complete, and hence $\Gamma(X,\mathcal O_X)$
is finite-dimensional over $K$. Indeed, since $X$ is also reduced, it is a reduced finite dimensional $K$-algebra. Since $X$ is furthermore irreducible,
it must be a field. (A reduced fin. dim'l. $K$-alg. is a product of fields,
but the product decomposition induces a corresponding decomposition of
$X$ into connected components. Since $X$ is irred. it is connected,
and so there can only be one such component.)
Thus the image of $A$ in $\Gamma(X,\mathcal O_X)$ is again a field, say $k$,
and so the morphism $X \to Y$ factors as $X \to $ Spec $k \to Y,$
and hence has as image the single closed point Spec $k$ of $Y$.