I heard from someone that, if $x:\text{Spec}k\to\mathcal{X}$ is a point of an Deligne-Mumford stack (algebraic should be OK, I am only assuming Deligne-Mumford so we know the residual gerbe exists), then the residual gerbe $\mathcal{Z}_x$ is étale-locally the classifying stack $\mathcal{B}_k\underline{ \text{Aut}}_k(x)$, i.e. the stack $[\text{Spec}k/\underline{ \text{Aut}}_k(x)]$, with the trivial action of $\underline{ \text{Aut}}_k(x)$, which is the automorphism group scheme of the point $x$.
The same person told me that if the residual gerbe is "split" or "neutralized", then in fact it is the classifying stack.
I cannot find a reference for the residual gerbe being étale-locally a classifying stack, nor can I even find a definition of what a "split" or "neutralized" gerbe is (perhaps they mean neutral? i.e. there is a section of $\mathcal{Z}_x \to Z$, where $Z$ is the algebraic space over which $\mathcal{Z}_x$ is a gerbe?) Sadly, I cannot simply ask this person, as they are not available for the next little while. Does anyone know anything about the above?