Let $V$ be a locally bounded measurable function such that $e^{-V}$ is bounded and belongs to $L^1(dx)$, and let $μ_V(dx):=C_V e^{-V(x)} dx$ be a probability measure. We present the criterion for the weighted Poincaré inequality of the non-local Dirichlet form $$ D_{ρ,V}(f,f):=\iint(f(y)-f(x))^2ρ(|x-y|) dy μ_V(dx) $$ on $L^2(μ_V)$. Taking $ρ(r)={e^{-δr}}{r^{-(d+α)}}$ with $0<α<2$ and $δ\geqslant 0$, we get some conclusions for general fractional Dirichlet forms, which can be regarded as a complement of our recent work Wang and Wang (2012), and an improvement of the main result in Mouhot, Russ, and Sire (2011). In this especial setting, concentration of measure for the standard Poincaré inequality is also derived. Our technique is based on the Lyapunov conditions for the associated truncated Dirichlet form, and it is considerably efficient for the weighted Poincaré inequality of the following non-local Dirichlet form $$ D_{ψ,V}(f,f):=\iint(f(y)-f(x))^2ψ(|x-y|) e^{-V(y)} dy e^{-V(x)} dx $$ on $L^2(μ_{2V})$, which is associated with symmetric Markov processes under Girsanov transform of pure jump type.