Let $\Omega$ be a bounded domain and $f_n\in L^2(\Omega)$ be a sequence such that $$\int_\Omega f_nq\operatorname{dx}\leq C<\infty\qquad \text{for all}\quad q\in H^1(\Omega),\ \|q\|_{H^1(\Omega)}\leq1,\ n\in\mathbb{N}.\quad (1) $$ Is it then possible to conclude that $$ \sup_{n\in\mathbb{N}}\|f_n\|_{L^2(\Omega)}\leq C. $$
Here, $H^1(\Omega)$ denotes the Sobolev-Hilbert-Space $H^1(\Omega)$.
Obviously, this statement would be true if we were to replace (1) with $$\int_\Omega f_nq\operatorname{dx}\leq C<\infty\qquad \text{for all}\quad q\in L^2(\Omega),\ \|q\|_{L^2(\Omega)}\leq1,\ n\in\mathbb{N}. $$
and maybe the dense and compact embedding $H^1(\Omega)\hookrightarrow L^2(\Omega)$ is of help but I'm not sure of it.
Edit: By now I'm pretty sure, that this statement doesn't hold. We only have a bound in the dual of $H^1(\Omega)$. But until now I'm failing to compile a conclusive argument!
Take $\Omega:=(0,1)$ and $f_n(x):=(x+1/n)^{-1/2}$. Then for all $q\in H^1(0,1)$ and all integer $n$, using continuity of the trace operator $\Gamma$, $$\int_0^1 f_n(x)q(x)dx=-2\int_0^1\sqrt{x+1/n}q'(x)dx+2\Gamma(q)\leqslant C\lVert q\rVert_{H^1}.$$ But $\lVert f_n\rVert_2^2=\int_0^1\frac 1{x+1/n}dx=\log(n+1)-\log n+\log n=\log(n+1)$.