The question is what 'scopes use the raw data for cursors and how can one tell the difference? Is it normally given in the specification?
The raw data is all you have.
Say a scope is at 2V/div, and there are 10 gridlines. In order to not the show the user a clipped signal, you'd want the input range to be +20V to -20V (not including any offset voltage the scope might be able to apply). That means 40V covered by an 8-bit ADC. That gives you 156 mV resolution.
On a good scope such as Rigol and beyond, you'd have a VGA after the input and as you reduced the input sensitivity, you'd increase the gain, so again your entire 8 bit could be used to digitize a +/- 10V signal, which would give 20/256=78mV resolution.
As you move down to 5mV/div, you'd again adjust the gain so that 8 bits was digitizing +/-50mV or 100/256 = 390uV resolution.
And of course, the analog offset voltage allows you still get that full 8 bit resolution on a signal riding atop a large DC component. For example, if I have a 30V DC signal, I can offset that with -30V and then set to 5mV/div and get 390uV resolution. Or switch to AC coupling.
Now, on a cheap scope, like a Link Instruments USB scope (MSO19), there isn't a variable gain amp after the input. So, as you get to lower voltage levels you can actually see the ADC quantiziation. But these are fortunately rare.
Thus, given a proper scope with a VGA following the input stage, you can practically measure just about any resolution signal you might want to know with great accuracy, regardless of scope ADC.
THE PRIMARY CASE where the an 8-bit ADC might bite you is if you are looking at noise and using your scopes FFT function.
An 8-bit scope will BEST CASE give you a noise floor of around -56 dB full scale at the highest sample rate. A 10-bit converter will do about 12 dB better. A 12-bit converter will be about 24 dB better than an 8 bit scope.
Note that fast sample rates can be traded off for resolution. A 10-bit scope running at 100MSPS that is sampling a signal at an effective 100KSPS can pick up almost 6 more bits of effective resolution, meaning a 10-bit converter can look (almost) like a 16-bit converter at the lower sample rates.