Dataflow gen 2 append option for data destination

When loading data to a lakehouse in Fabric using a Gen 2 dataflow, how exactly does the append option work?

Does it only do append operation by inserting rows with previously unseen primary keys or does it also update existing rows with changes from the source?

I have an on prem SQL server which I'm trying to ETL data from into a lake house and it's unclear if the append option is a pure append or if it is analagous to an "upsert"