Yes it definitely WILL affect the reading!
High impedance voltage source and high sampling frequencies equal distorted wrong ADC readouts. The current, flowing due to the sampling capacitor being periodically charged will draw current from the divider, producing error in the readout.
Note that using a decoupling capacitor on the output of the divider will ONLY help with a single-shot measurement, it won't fix the current drawn from the source (due to I = Csample * Usample * fsample).
And also, the sampling capacitor is non-linear, or the drawn current respectively: The ADC may both inject current or draw a current from the source, depending on the instantaneous voltage present (the internal workings of the ADC are not so trivial as they teach in the school).
It is always a good practice to buffer ADC inputs with a "zero impedance" buffers, with additional decoupling at the ADC input pin to compensate for the pulsed current from the sampling action.
For 1MSps, there is no way a 9k1 - 1k0 divider would be sufficient. You need a low impedance source. Say 100ohm impedance.