In a paper by Heath-Brown, upon having to estimate the number of solutions $(x,y)\in\mathbb Z^2\cap[-B,B]^2$ to the equation $Q(x,y)=k$ with $Q$ an integer-coefficient non-degenerate quadratic form and $k$ an integer, he simply states that by "standard facts from the theory of quadratic forms" this is $O_\varepsilon(H(Q)^\varepsilon B^\varepsilon)$, where $H(Q)$ is the maximum absolute value of $Q$'s coefficients.
I have tried searching for proofs of this online, but no source mentions any inequality of the sort. My attempts at proving it have been mostly analytical, notably by attempting to estimate the integral of the generating function $$S(\alpha,B)=\sum_{(x,y)\in\mathbb Z^2\cap[-B,B]^2}e(\alpha Q(x,y))$$ like in the circle method but nothing has come of it.
Also, his comment on "standard facts" make me think this must be an algebraic fact rather than an analytical one, but trying to limit the values of $x$ and $y$ to divisibility conditions involving $k$ haven't brought me any further.
Any help or reference would be appreciated. Thanks !