The good people at Apple, knowing that folk will merrily plug the headphone jack into all kinds of speakers probably already included the capacitor in series to get rid of any DC offset. At least that's what I'd do if I was in charge....
@Aubrey: If you want to get rid of the DC component of the signal just use a capacitor in series with the signal.
Absolutally -- in fact the circuit I'm using with the MSGEQ7 chips from SparkFun has just such a decoupling capacitor (I'll be discussing thsi circuit in a future column) -- the main thing at this point is that I wanted to use my frequency generator to replicate the signal coming out of the iPad -- if there had been a DC offset I woudl have replicated it, even though my series capacitor would have (effectively) taken it out again.
If you want to get rid of the DC component of the signal just use a capacitor in series with the signal. On t'other side you need a largish resistor to ground if you are looking to ground reference it, but you can bias it to whatever DC value you like by using a resistor divider.
Remember to choose the value of the capacitor for a low impedance and that it will interrelate with the aformentioned resistors. Today you can get ceramic caps in the 10uF range or more, but in the old days you had to use polarized ones leading you to try and figure out which side should be positive, Of course you could depolarize by placing two in series in opposing polarity.
I am surprised there wasn't a DC bias. From my understanding (which is very little) of a TI seminar on the audio amplifiers in phones etc. the voltage swing is doubled by driving the output in opposite phases across the speaker, much like the driver on RS485 and as I recall it ran on a single supply. Obviously I misunderstood.