Domain Models are only for the "write" side

A Domain Model is relevant only when changing the state of a system. Aggregate boundaries, business invariants, and services are modeled to protect and preserve the state of the system. They are not relevant and don't play a role on data fetch.

Consequently, the "read" side of the application can be a completely different component/service, without having any linkage to the Domain Model. Indeed, the strength of CQRS design patterns is this clear segregation between the write and the read sides of the application.

The read models are completely delinked from aggregate structures and are custom built to serve a specific need, like an API response or a Business Report. Usually, there are many representations of the same data, in the form of multiple read models.

Also, as a consequence, the structure of persisted data no longer matters. Data can be stored in any format that is deemed sufficient, from simples text files to events in an Event store.