In simplistic terms, the signal has a fixed carrier signal and two side lobes which contain the audio info
If using an 8.33 spacing, that carrier needs to be in the middle of that spacing, or its side lobes will affect other channels.
If using a 25kHz spacing, there is more flexibility, - that carrier can be within a wide area in that spacing.
The side lobes are mirrors of each other, so to get the frequency bandwidth required to transmit sound, you need to multiply by 2 the maximum allowed audio frequency.
So 8.33kHz spacing means the maximum audio frequency possible is 8.33/2 = 4.12kHz. Generally the filter is 7kHz wide, so only 3.5kHz.
Human voice spectrum:
But with the key vowels, which makes speech intelligible here:
So its possible to cut off some of the top of the full spectrum and still have intelligible words. Music might not sound good, and a high soprano singer might get cut off, so perhaps listening to classical music over airband wouldn't be the best, but speech still works.
The 25kHz spacing allows for old transceivers to not be accurate with their transmitting frequency yet still be picked up.
It is also used by some wide area services (D&D, Scottish Info, London Info, Farnborough North etc) - they transmit the same thing at different points within that spacing, far enough apart that the side lobes wont interact. Eg Scottish info's Craigowlhill TX/RX site operates on the spot frequency of 119.875mhz, Lowther hill operates on an offset of -7.5khz and Windy head on an offset of + 7.5khz. So all sites transmit within the same 25kHz frequency, but if you're high enough up to pickup multiple sites, they shouldn't interfere.
So your radio must be able to switch receive filter between 8.33kHz spacing and 25 kHz spacing, or you wouldn't be able to pickup the wide area transmitters or you will get interference between different 8.33 kHz channels.
Does all of that make any sense?