The idea of leaky abstractions seems relevant here. This is the observation from engineering that when layers of models are built on top of each other, consequences of the lower-level models tend to appear even when the higher layers are meant to abstract them away.

Asking whether a model is "real" seems akin to asking whether its abstraction will ever leak, and if it does, whether the places where lower layers show through are correctly labelled and explained within the model. Atomic chemistry is "real" in that, when it does break down (extreme energies, rare particles, etc), it's for reasons that can be explained in atomic chemistry's own vocabulary. On the other hand, psychology, for example, tends to break down for reasons that can't be explained, or can only be explained in terms of biology.

Under this definition, if the universe is a simulation, then it is real if and only if that simulation runs to completion without information about the simulator's universe leaking into our universe.