Description:
If you want to switch from standard MySQL replication to GTID replication, you must shut all your servers down at the same time.
This means downtime, which can be significant if you have a large number of replicas: this is bad!
How to repeat:
1/ Set up one master and one slave (both 5.6) using standard replication.
2/ Try to switch to GTID replication without stopping both servers and without breaking replication.
Suggested fix:
I think you should be able to restart a server with gtid_mode=ON and keep using replication with binlog file/binlog pos.
That would allow you to restart all your slaves one by one (no downtime), fail over to one of the slave to be able to restart the master (small downtime), fail back to the old master (again small downtime) and then enable GTID replication on all slaves, one by one (no downtime).