There are both technical and biological reasons to do GC-content normalization.

Technically, both PCR efficiency of target sequences, and binding of probes to target sequences are related to GC content of the sequence. Especially without normal samples this is a problem.

The technical issues could be solved by testing paired normal sampled, but we still see copy number effects that show a relationship with GC-content. It appears that replication timing during the cell cycle is related to GC-content. Because the growth-fraction in tumor tissue is generally higher than in normal tissue, these effects can be seen in copy-number profiles. GC-content is easier to obtain than replication timing, which is why generally GC-content normalization is used.