There's nothing stupid with your post, as your ADC is still in danger!

If you connect analog circuitry powered by relevantly high bipolar supply voltage to +5V powered ADC chip, then you are always running into danger of heavily overranging the ADC! Why?
Because there are conditions thinkable, where either output voltage of OPamp delivers much more than +5V, or where accidantly even negative voltages can be present at input of ADC. Think of power-on and power-off procedures, for instance.

Sometimes, it's not trivial to find a properly working protection scheme for the ADC...

Let's have a look at datasheet of PCF8591:
Input voltages must be lower than Vcc + 0.5V and higher than -0.5V. This makes us believe, that from each input to supply rails protection diodes are implemented on chip. But these 0.5 volts, which are typical for a forward biased diode, need not necessarily to be the voltage drop of a protection diode!!! It can also be an indication, that a latch-up causing parasitic thyristor is present, which will be fired at input voltages exceeding given values!
So, we search for a maximum input current specification, which shows us, what input currents are allowed, which

either do not destroy protection diodes (if they are such!)

or do not fire internal parasitic thyristor.

We find, that input current must be smaller than +-10mA. Voila!

As input of your ADC is very high ohmic (which is not all always the case with ADCs!), we are allowed to insert a simple series resistor directly from output of OPamp to input of ADC to provide a current limiting. Looking at datasheet will make us believe, that introducing such a resistor is a good idea, because typical specifications assume a source resistance of Rs = 10kOhm!
So, if we want to limit input current into ADC to a sane level, we must choose in your case: R >= 9V / 10mA = 900Ohm. So, everything between 1k and 10k should do the trick.

Kai