Incremental syncing involves transferring only new or updated data, thus avoiding duplication of previously replicated data. This is achieved through deduplication using a unique primary key specified in the model. For initial syncs, it functions like a full refresh since all data is considered new.