As Mike has pointed out, the main errors come from charging and discharging the internal sample and hold capacitor, as long as you allow time for that to happen it's not at all critical.
Also, if you're only using one ADC input, and it's permanently enabled, there's little problem - it's switching between them that causes the main errors.
Try an experiment:
Write a program to display two ADC inputs on an LCD, switching as fast as you can between the inputs (allowing the minimum conversion time).
Connect the sliders of two 4.7K pots to each input, and the other ends to Vss and Vdd.
Now alter the pots while you watch the display, each should only alter it's own reading.
Now add 100K resistors between the ADC's and the sliders, and repeat.
You'll now see that each pot affects both readings - a LOT.
This is because the capacitor doesn't have time to charge/discharge.
Now add longer time delays between selecting the input and taking the reading, make them long enough and the pots won't interact any more.