Head desking at an infrequent sight on Vox sites

gistfile1.txt

Unicode is what we all use* to store our characters. Because of legacy reasons,
we needed a way to store them so that the basics mapped onto the old system
(ASCII) so UTF-8 was created. It is a representation of Unicode with certain
constraints on how we store our characters. For something like an em dash ( — ),
this is U+2014 in Unicode, and represented as
0xE2 0x80 0x94 in UTF‐8 bytes.

What I assume is happening in this case, is your writers are using em dashes
freely in their writing, copy/pasting the content into your CMS and at this
point it is converting the above sequence into:
0xC2 0xE2 0xC2 0x80 0xC2 0x94
which would show to the user: â[][] (most platforms will hide control
characters like these)

At this point they probably just search & replace for â → — which will clean
the content for them, but the source still retains the other characters the CMS
mung, so now it is stored as:
0xE2 0x80 0x94 0xC2 0x80 0xC2 0x94
which is how it is showing for me: http://i.imgur.com/MgTRmfS.png

* This masks lots of history and culture and I am sure you can Wikipedia
  character sets and the history of Operating Systems to learn more if you need
  to