Used fiddler to double-check* this.
Everything I get back from the API is definitely gzip compressed.
For that matter, you can't get a response that isn't compressed; you can only choose between gzip & deflate. Default is gzip.
The API always sets the Content-Encoding response header, so you might be the victim of your library decompressing the response for you.
*Triple-checked: I ran our test cases without setting Accept-Encoding but leaving the gzip decompression code in.