In addition to a little reverb (convolution reverbs are critical tools for sound for picture, IMHO), it will still sound like soundtrack music with extra reverb unless it is filtered.
You could run it through a speaker emulator or guitar or bass amp plug-in, or you could do high and low pass filters along with a little high shelf. Instead of low pass + high shelf, you could do a 3 dB/octave or otherwise more shallow sloped-low pass.
For a loud dance club, the high pass could be left off to keep the sense of thumping bass, but the low pass/high shelf is still important. A sidechain compressor to gently duck the music under the dialog wouldn't hurt either and would help make the music sound louder without stomping too much on the dialog.
Make sure you filter/EQ before the reverb.
Keep in mind that high frequencies are attenuated more quickly than low frequencies as one moves farther away from the sound source. See this Physics.SE question on that topic. That means we can make things sound farther away by reducing the high frequencies. Also, many speaker systems usually top out around 16 kHz at best, and high frequencies do not travel through intervening walls or diffract around corners as well as lows. Going back to the shallow-sloped low pass, you can move a sound farther away just by slowing dialing down the corner frequency.
Finally, remember that sound design serves the picture, it doesn't rule it. Do whatever you have to do to make the sound get the point of the scene across. If there's heartfelt dialog going on, the music better not stomp all over it, so no matter how unrealistic or crazy the final sound is (ducking doesn't happen in real life), it's better to do whatever needs to be done to make the sound work.