Proof. Note first that a graph $G$ in $G(n,p)$ has diameter $1$ iff it's a complete graph in $n$ vertices, so that $P(\diam G=1)=p^{\binom{n}{2}}\to 0$ when $n\to\infty$. So, if we prove that almost every graph in $G(n,p)$ has diameter $\leq 2$ we're done.
Suppose a graph $G$ has diameter $>2$. Then there are two vertices $u,v$ such that $d(u,v)>2$, so that $N(u)\cap N(v)=\varnothing$. Conversely, if this last property holds for some graph, the graph must indeed have diameter $>2$. So, to prove that almost every graph in $G(n,p)$ has diameter $2$ we need to prove that almost every graph in $G(n,p)$ does not have vertices with disjoint neighborhoods. Consider the following random variable on $G(n,p)$:
$$X(G)=|S(G)|$$
By Markov's inequality,
$$P(X\geq 1)\leq E(X),$$
so we're left to prove that $\lim_{n\to\infty} E(X)=0$.
Now, $E(X)=\sum_{G\in G(n,p)}P(G)X(G)$.
For a given vertex $x\in V(G)$, the probability of it being adjacent to both vertices $u,v$ is $p^2$, so the probability of $x\notin N(u)\cap N(v)$ is $1-p^2$, so $P(N(u)\cap N(v)=\varnothing)=(1-p^2)^{n-2}$. Also $P(uv\notin E(G))=1-p$.
So,
$$E(X)=\binom{n}{2} (1-p)(1-p^2)^{n-2}.$$
Letting $n\to\infty$, since $|1-p^2|<1$, $(1-p^2)^{n-2}\to 0$ and $\lim_{n\to \infty} E(X)=0$.
Thus $\lim_{p\to \infty} P(X\geq 1)=0$, so that almost every graph in $G(n,p)$ has no such vertices $u,v$, and therefore almost every graph in $G(n,p)$ has diameter $\leq 2$, and still almost every such graph has diameter $2$.