The following theorem can be found in Milne's book on algebraic groups (Theorem 1.45) and in his online course notes (Proposition 1.31):
Theorem: Let $G$ be an algebraic group (= algebraic group scheme, not necessarily affine) over a field $ k $ and $S$ a closed subgroup of $G(k)$. Then there is a unique reduced algebraic subgroup $H$ of $G$ such that $ H(k) = S$.
From this I deduce:
Corollary: Let $G$ be a reduced algebraic group over a field $ k $ and let $H$ be a reduced algebraic subgroup of $G$ such that $ H(k) = G(k) $. Then $ H = G $.
Proof: Apply the Theorem with $ S = G(k) $.
However:
Counterexample to the Corollary: Let $k$ be the field of real numbers and let $G=\mu_3$ be the group of 3rd roots of unity (i.e. the Hopf algebra of $G$ is $k[X]/(X^3-1)$). Since $k$ is of characteristic zero, $G$ is smooth and thus reduced. Let $ H $ be the trivial subgroup of $G$ (i.e. $H(R) = \{1\} $ for all $k$-algebras $R$). Then $ H(k) = \{1\} = G(k) $ and $G,H$ are both reduced, but $G \neq H$.
Where is my mistake?
(If it helps, I only care about affine group schemes.)