[step:Bound the rank of an interpreted name by the rank of the name]Let $\tau \in M$ be a $\mathbb{P}$-name, meaning a hereditary collection of pairs $(\sigma,p)$ with $\sigma$ again a $\mathbb{P}$-name and $p \in \mathbb P$. Define the $M$-rank of $\tau$, denoted $\operatorname{rank}^{M}(\tau)$, to be the ordinal computed in $M$ by the recursive rule
\begin{align*}
\operatorname{rank}^{M}(\tau)=\sup \{\operatorname{rank}^{M}(\sigma)+1 : \exists p \in \mathbb{P}\, ((\sigma,p) \in \tau)\}.
\end{align*}
Since $M \models \mathrm{ZFC}$, this rank is an ordinal of $M$. Because $M$ is transitive and the recursion defining name rank uses only membership among names already in $M$, this ordinal agrees with the ambient recursive rank of the same name. For any set $y \in M[G]$, write $\operatorname{rank}(y)$ for the ordinary von Neumann rank of $y$ computed in the transitive model $M[G]$, equivalently the ambient rank of $y$ in the surrounding universe.
Here $\tau_G$ denotes the name evaluation of $\tau$ by $G$, namely the set obtained recursively by keeping exactly the interpretations $\sigma_G$ for pairs $(\sigma,p) \in \tau$ with $p \in G$. We claim that for every $\mathbb{P}$-name $\tau \in M$,
\begin{align*}
\operatorname{rank}(\tau_G) \leq \operatorname{rank}^{M}(\tau).
\end{align*}
[claim:Evaluation does not increase rank beyond the name rank]
For every $\mathbb{P}$-name $\tau \in M$, the interpreted set $\tau_G$ satisfies
\begin{align*}
\operatorname{rank}(\tau_G) \leq \operatorname{rank}^{M}(\tau).
\end{align*}
[/claim]
[proof]
We prove the claim by induction on the ordinal $\operatorname{rank}^{M}(\tau)$. Fix a $\mathbb{P}$-name $\tau \in M$, and assume the assertion has already been proved for every $\mathbb{P}$-name $\sigma \in M$ with
\begin{align*}
\operatorname{rank}^{M}(\sigma) < \operatorname{rank}^{M}(\tau).
\end{align*}
By definition of name evaluation,
\begin{align*}
\tau_G=\{\sigma_G : \exists p \in G\, ((\sigma,p) \in \tau)\}.
\end{align*}
Let $x \in \tau_G$. Then there exist a $\mathbb{P}$-name $\sigma \in M$ and a condition $p \in G$ such that $(\sigma,p) \in \tau$ and $x=\sigma_G$. From the definition of $\operatorname{rank}^{M}(\tau)$, this implies
\begin{align*}
\operatorname{rank}^{M}(\sigma) + 1 \leq \operatorname{rank}^{M}(\tau),
\end{align*}
and hence
\begin{align*}
\operatorname{rank}^{M}(\sigma) < \operatorname{rank}^{M}(\tau).
\end{align*}
The induction hypothesis gives
\begin{align*}
\operatorname{rank}(x)=\operatorname{rank}(\sigma_G)\leq\operatorname{rank}^{M}(\sigma)<\operatorname{rank}^{M}(\tau).
\end{align*}
Since this holds for every $x \in \tau_G$, the definition of rank gives
\begin{align*}
\operatorname{rank}(\tau_G)=\sup\{\operatorname{rank}(x)+1 : x \in \tau_G\}\leq\operatorname{rank}^{M}(\tau).
\end{align*}
This proves the claim.
[/proof][/step]