What version is your Atlas cluster? We do offer different read Preferences for your operations as part of the client logic.
Now there are a few considerations you need to be aware of:
The secondary is going through same write workload as the primary. Non blocking reads (blocked behind replication) were introduced only in 4.0+. So although you will ease your primary you will not necessarily get better performance
Since replication is asynchronous you may read stale data. Not all reads can live with that.
The driver does not necessarily load balance connections across secondaries and therefore it will be more a round robin assignment.
I would suggest that if you have specific load you want to offload from the primary for analytics you should provision ANALYTIC nodes .
MongoDB official stand on scaling is by using sharding as the best scale out option. You can always just scale up with raising the Atlas tier as well.
Please read more on our production consideration: