Does embed_content return normalize embedding after reducing output_dimensionality

For “semantic_similarity” use case if I use the output_dimensionality to reduce down to 256, does the API automatically normalize?

Update, I found that you can test for normalization and found that it does not return it normalized.