[guided]For each $t \geq 0$, define the characteristic map $F_t: \mathbb{R} \to \mathbb{R}$ by
\begin{align*}
F_t(a) = a + t u_0(a).
\end{align*}
The characteristic formula gives the label $a$ implicitly through
\begin{align*}
x = F_t(a) = a + t u_0(a).
\end{align*}
On a relatively open spacetime region $\Omega \subset [0,\infty) \times \mathbb{R}$ where this equation can be solved by a $C^1$ local inverse branch $A: \Omega \to \mathbb{R}$, the value $A(t,x)$ is the characteristic label that reaches the spacetime point $(t,x)$, and it satisfies
\begin{align*}
A(t,x) + t u_0(A(t,x)) = x.
\end{align*}
The differentiations below are ordinary partial derivatives at points with $t>0$; if the initial slice is included, the time derivative at $t=0$ is understood as a right derivative.
We then define the candidate solution by transporting the initial value along the characteristic:
$u: \Omega \to \mathbb{R}$ by
\begin{align*}
u(t,x) = u_0(A(t,x)).
\end{align*}
In this proof, a classical solution means a function $u \in C^1(\Omega)$ satisfying the differential equation pointwise at interior points with $t>0$ and satisfying the initial condition on the initial slice where that slice is included in $\Omega$.
The key quantity controlling the inverse branch is the derivative of the characteristic map with respect to the label. Define $J: \Omega \to \mathbb{R}$ by
\begin{align*}
J(t,x) = 1 + t u_0'(A(t,x)).
\end{align*}
We now compute the derivatives of $A$ from the identity
\begin{align*}
A(t,x) + t u_0(A(t,x)) = x.
\end{align*}
Differentiating with respect to $x$ gives
\begin{align*}
\partial_x A(t,x) + t u_0'(A(t,x))\partial_x A(t,x) = 1.
\end{align*}
Factoring the left-hand side gives
\begin{align*}
J(t,x)\partial_x A(t,x) = 1,
\end{align*}
so $J(t,x) \neq 0$ at every point of the branch and
\begin{align*}
\partial_x A(t,x) = \frac{1}{J(t,x)}.
\end{align*}
Differentiating the same identity with respect to $t$ gives
\begin{align*}
\partial_t A(t,x) + u_0(A(t,x)) + t u_0'(A(t,x))\partial_t A(t,x) = 0.
\end{align*}
Again factoring the terms involving $\partial_t A$ gives
\begin{align*}
J(t,x)\partial_t A(t,x) + u_0(A(t,x)) = 0,
\end{align*}
and therefore
\begin{align*}
\partial_t A(t,x) = -\frac{u_0(A(t,x))}{J(t,x)}.
\end{align*}
Now apply the chain rule to $u(t,x)=u_0(A(t,x))$. We obtain
\begin{align*}
\partial_t u(t,x) = u_0'(A(t,x))\partial_t A(t,x)
\end{align*}
and
\begin{align*}
\partial_x u(t,x) = u_0'(A(t,x))\partial_x A(t,x).
\end{align*}
Substituting the formulas for $\partial_t A$ and $\partial_x A$ gives
\begin{align*}
\partial_t u(t,x) = -\frac{u_0'(A(t,x))u_0(A(t,x))}{J(t,x)}
\end{align*}
and
\begin{align*}
u(t,x)\partial_x u(t,x)= u_0(A(t,x))\frac{u_0'(A(t,x))}{J(t,x)}.
\end{align*}
Adding these two expressions gives
\begin{align*}
\partial_t u(t,x) + u(t,x)\partial_x u(t,x) = 0.
\end{align*}
Thus the parametrisation produces a classical solution wherever the inverse branch exists, with the one-sided time derivative convention on the initial boundary. At $t=0$, $F_0(a)=a$, so the inverse label is $A(0,x)=x$ and the initial condition becomes
\begin{align*}
u(0,x)=u_0(A(0,x))=u_0(x).
\end{align*}[/guided]