In QED, the electromagnetic (EM) field and the charged matter
are both quantum entities.
This answer uses a semiclassical model instead,
with a quantum field coupled to a prescribed classical current.
This is an exactly solvable model inspired by QED.
As a further simplification, the quantum field
will be a scalar field instead of the EM field.
By analogy, the quanta of this scalar field will be called "photons".
In the context of
the free (non-interacting) quantum EM field,
the word "photon" is typically used
to mean a quantum of energy,
and that's how I'm using the word here.
The current will be active only during a finite time interval,
and I'll only apply the word "photon"
at the times when the current is not active,
so that the meaning of "quantum of energy" is unambiguous.
To help limit the length of this post,
familiarity with introductory QFT is assumed.
The notation will be similar to that used in
chapter 2 of Peskin and Schroeder's
An Introduction to Quantum Field Theory.
The model and its exact solution
The Heisenberg picture will be used, so the state-vector
is independent of time, but its physical significance
still changes in time because the observables do.
The equation of motion in the Heisenberg picture is
\begin{equation}
\partial_\mu\partial^\mu\phi(t,\mathbf{x})
= J(t,\mathbf{x})
\tag{1}
\end{equation}
where $\phi$ is the quantum field
and where $J$ is a prescribed function that will
be called the "current" by analogy with the EM case.
The equal-time commutation relation for the quantum scalar
field is
\begin{equation}
\big[\phi(t,\mathbf{x}),\,\dot\phi(t,\mathbf{y})\big]=i\delta^3(\mathbf{x}-\mathbf{y}).
\tag{2}
\end{equation}
The quantum field $\phi(t,\mathbf{x})$ is the local observable
corresponding to field-amplitude measurements.
Equations (1)-(2) can be solved exactly.
The solution is
\begin{equation}
\phi(t,\mathbf{x}) = \phi_0(t,\mathbf{x})+\phi_J(t,\mathbf{x})
\tag{3}
\end{equation}
where:
$\phi_J$ is a real-valued solution to (1), which commutes with
everything;
$\phi_0$ is an operator-valued solution to the $J=0$ version of (1)
that satisfies the commutation relation (2).
From now on, suppose the current is non-zero only within
the finite time interval $0<t<T$:
\begin{equation}
J(t,\mathbf{x})=0
\hskip1cm
\text{ except for }0<t<T
\tag{4}
\end{equation}
and choose
\begin{equation}
\phi_J(t,\mathbf{x})=0
\hskip1cm
\text{ for }t\leq 0.
\tag{5}
\end{equation}
These conditions are all satisfied by
\begin{equation}
\phi_0(t,\mathbf{x})
=\int\frac{d^3p}{(2\pi)^3}\ e^{i\mathbf{p}\cdot \mathbf{x}}\,
\frac{e^{-i\omega t}a_0(\mathbf{p})+e^{i\omega
t}a_0^\dagger(-\mathbf{p})}{
\sqrt{2\omega}}
\tag{6}
\end{equation}
and
\begin{equation}
\phi_J(t,\mathbf{x})
= \int ds\ \theta(t-s)
\int\frac{d^3p}{(2\pi)^3}\ e^{i\mathbf{p}\cdot \mathbf{x}}\,
i\,\frac{e^{-i\omega (t-s)}-e^{i\omega (t-s)}}{2\omega}\,
\tilde J(s,\mathbf{p})
\tag{7}
\end{equation}
with
\begin{equation}
\omega\equiv \sqrt{\mathbf{p}^2}
\hskip2cm
\tilde J(s,\mathbf{p})
\equiv
\int d^3x\ e^{-i\mathbf{p}\cdot \mathbf{x}}\,J(s,\mathbf{x}),
\tag{8}
\end{equation}
and where the operators $a_0(\mathbf{p})$ and their adjoints satisfy
\begin{equation}
\big[a_0(\mathbf{p}),\,a_0^\dagger(\mathbf{p}')\big]=(2\pi)^3\delta^3(\mathbf{p}-\mathbf{p}').
\tag{9}
\end{equation}
The operators $a_0$ and $a_0^\dagger$ are
just a basic set of operators in terms of which everything
else in the operator algebra may be expressed.
Define a state-vector $|0\rangle$ by the conditions
\begin{equation}
a_0(\mathbf{p})\,|0\rangle = 0
\hskip2cm
\langle 0|0\rangle = 1
\tag{10}
\end{equation}
for all $\mathbf{p}$,
and suppose that the state of the system
is the one represented by $|0\rangle$.
The Heisenberg picture is being used here,
so the state-vector has no time-dependence,
but the physical state that it represents
still changes in time because the observables do.
The rest of this answer addresses the interpretation
of the state-vector (10) both for $t<0$ and for $t>T$,
first in terms of photons and then as it relates to radio waves.
The interpretation in terms of photons
Equation (5) says that for $t<0$ we have the familiar free scalar
field,
and then we recognize the state defined by (10) as the vacuum state
—
the state of lowest energy, with no photons.
This, of course, was the motive for choosing the state (10).
The question is what happens at $t > T$
in the aftermath of the temporary current $J$.
For these times, equation (4) says that the factor
$\theta(t-s)$ may be omitted in equation (7),
because it is already enforced by the current itself.
Therefore, for these late times, the solution (3) may be written
\begin{equation}
\phi(t,\mathbf{x})
=\int\frac{d^3p}{(2\pi)^3}\ e^{i\mathbf{p}\cdot \mathbf{x}}\,
\frac{e^{-i\omega t}a(\mathbf{p})+e^{i\omega
t}a^\dagger(-\mathbf{p})}{
\sqrt{2\omega}}
\tag{11}
\end{equation}
with
\begin{equation}
a(\mathbf{p})
\equiv a_0(\mathbf{p}) + a_J(\mathbf{p})
\hskip2cm
a_J(\mathbf{p}) \equiv \frac{i}{\sqrt{2\omega}}\int ds\
e^{i\omega s}\tilde J(s,\mathbf{p}).
\tag{12}
\end{equation}
The complex-valued function $a_J$ encodes the effect of the current.
Before we can interpret the state (10) in terms of photons at times
$t>T$,
we need to determine which operators represent photon
creation/annihilation operators at these times.
The Hamiltonian associated with the equation of motion (1) is
\begin{equation}
H(t)
=
\int d^3x\ \left(\frac{\dot\phi^2(t,\mathbf{x})
+(\nabla\phi(t,\mathbf{x}))^2}{2}-\phi(t,\mathbf{x})
J(t,\mathbf{x})\right).
\tag{13}
\end{equation}
Equations (9) and (12) imply
\begin{equation}
\big[a(\mathbf{p}),\,a^\dagger(\mathbf{p}')\big]
=(2\pi)^3\delta^3(\mathbf{p}-\mathbf{p}').
\tag{14}
\end{equation}
Any any time for which $J=0$, equations (11) and (13)-(14) imply
\begin{align}
H(t)=
\int\frac{d^3p}{(2\pi)^3}\ \omega\,a^\dagger(\mathbf{p})a(\mathbf{p})
+h(t)
\tag{15}
\end{align}
where $h(t)$ is a real-valued function that doesn't affect this
analysis.
Whenever $J=0$, these equations all have the same form as they
do in the free-field case (where $J$ is zero for all times).
Based on this, we can interpret $a(\mathbf{p})$
and its adjoint as the operators that annihiliate and create
(respectively) a photon with the indicated momentum at times $t>T$.
The justification for this interpretation
is identical to the corresponding justification for $a_0$ at times
$t<0$.
Now that we know which operators create and annihilate photons
at $t>T$, we can interpret the state $|0\rangle$ at these times.
Equations (10) and (12) imply
\begin{equation}
a(\mathbf{p})\,|0\rangle = a_J(\mathbf{p})\,|0\rangle,
\tag{16}
\end{equation}
which is the defining equation of a multi-mode coherent state.
The state $|0\rangle$ was chosen because
it represents the vacuum state for $t<0$,
but equation (16) says that it is no longer the vacuum state with
respect
to observables at $t>T$.
The vacuum state at $t>T$ is represented instead by the state-vector
$|T\rangle$ that satisfies
\begin{equation}
a(\mathbf{p})\,|T\rangle = 0.
\tag{17}
\end{equation}
Equation (14) implies that
the relationship between the coherent state (16) and the vacuum state
(17)
is
\begin{equation}
|0\rangle \propto \exp\big(A^\dagger \big)\,|T\rangle
=|T\rangle + A^\dagger|T\rangle
+ \frac{1}{2!}(A^\dagger)^2|T\rangle
+ \frac{1}{3!}(A^\dagger)^3|T\rangle
+\cdots
\tag{18}
\end{equation}
with
\begin{equation}
A^\dagger \equiv
\int\frac{d^3p}{(2\pi)^3}\
a_J(\mathbf{p}) a^\dagger(\mathbf{p}).
\tag{19}
\end{equation}
In words, the state at times $t>T$ is
a special superposition of different numbers of
identical photons, all with this same profile described by the complex-valued function $a_J(\mathbf{p})$.
The interpretation as a radio wave
At any time $t$, equations (3)-(10) imply
\begin{equation}
\langle 0|\phi(t,\mathbf{x})|0\rangle=\phi_J(t,\mathbf{x})
\tag{20}
\end{equation}
and
\begin{equation}
\langle 0|\phi(t,\mathbf{x})\phi(t,\mathbf{y})|0\rangle -
\langle 0|\phi(t,\mathbf{x})|0\rangle\,
\langle 0|\phi(t,\mathbf{y})|0\rangle
=
\langle 0|\phi_0(t,\mathbf{x})\phi_0(t,\mathbf{y})|0\rangle.
\tag{21}
\end{equation}
Equation (20) says that the expectation value
of the quantum field behaves just like a classical wave
generated by the current $J(t,\mathbf{x})$.
Equation (21) says that
the fluctuations in the outcomes of field-amplitude measurements
are just as small as they would be in the vacuum.
If the current $J$ is large enough, so that the expectation value (20)
is large enough, then the square root of (21) will be negligible
compared to (20).
In this case, we have a classical wave for all practical
purposes.
By choosing the oscillation frequency
of the current, we can make it a radio wave.
Altogether, this shows that if we start with the vacuum at
time $t<0$ and turn on a current during the interval $0<t<T$,
then the state at times $t>T$ is a coherent state of photons,
and the same state
can also be interpreted as an effectively classical wave.
Classical superposition versus quantum superposition
Note that a classical superposition of two such effectively-classical
waves is obtained by adding the corresponding
single-photon profiles in the
exponent of equation (18), like this:
\begin{equation}
\exp\big(A_1^\dagger + A_2^\dagger\big)\,|T\rangle.
\tag{22}
\end{equation}
This follows from the fact that such a superposition is produced by a
classical current of the form $J=J_1+J_2$, where $J_1$ and $J_2$
may be localized in different regions of space (for example).
In contrast, a quantum superposition of two
effectively-classical waves has the form
\begin{equation}
\exp\big(A_1^\dagger\big)\,|T\rangle +
\exp\big(A_2^\dagger\big)\,|T\rangle.
\tag{23}
\end{equation}
In this state, equation (21) doesn't hold;
fluctuations in field-amplitude measurement-outcomes
are typically just as large as the expectation value,
so a quantum superposition of two effectively-classical
waves not like a classical wave at all.
