Since the form takes values in the prime field it is nigh impossible for it to vanish only when $a=b$. Indeed, for a given $a$ it vanishes for at least half of the possible values of $b$ (if not all) by bilinearity alone. Hence I assume that the question is really to prove that the
bilinear form vanishes whenever $a=b$.
I give two proofs. The latter is more trivial, but the former is suggested by
my answer to another question about this form. There I derive the formula
$$
B(a,b)=tr(ab)+tr(a)tr(b).
$$
Using the fact $tr(a)=tr(a^2)$ (the Frobenius conjugates have the same trace) we can deduce the present claim easily.
If $a=b$ we have
$$
B(a,a)=tr(a^2)+tr(a)^2=tr(a)+tr(a)^2.
$$
As $tr(a)$ is either $0$ or $1$ this is equal to zero.
Oh dear, I'm sluggish. Here is a straightforward proof:
If $a=b$, then $a^{2^i}b+ab^{2^i}=a^{1+2^i}+a^{1+2^i}=0$ for all $a$ and all $i$.
The claim follows immediately from this.