The best way to get 50 ohms across the full 2.4GHz ISM band is to use a 50 ohm resistor, but it won't radiate very well, at least not at 2.4GHz. That's why it's called a dummy load. But it is broadband.
Another way to get close is to use a 1/4 wave ground plane in free air, characteristic impedance not a million miles away from 50 ohms.
Designing antennas is one of those wonderful areas of engineering where there are a number of parameters that are not mutually exclusive, they interact, and you have to come up with a compromise.
Here's a list off the top of my head...
a) Impedance
b) Radiation pattern
c) Bandwidth
d) Efficiency
e) Size
You simply won't get an antenna that's 50 ohm resistive across the band, unless it's a resistor. The most efficient antennas are resonant, and so have a limited bandwidth. You can increase bandwidth, but it'll be at the expense of efficiency and/or size.
So you have to choose the order of your priorities are design around that.
By far my best recommendation from what you've said so far is to find a generally acceptable proven PCB design, and go with that. You can model antennas, but again, it's like the squidgy balloon metaphor, you fix one parameter and that messes up with one or two other ones. Far easier, unless you want to spend months if not years of your life learning about the various intricacies and voodoo of antenna design.
However, if you do, Antennas by John Kraus is the generally accepted seminal work to start with. More applicable to what you're doing is the Microstrip Antenna Design Handbook or Microstrip and Printed Antennas: New Trends, Techniques and Applications, I have the former, it's a very dull read, you've really got to be desperate, I guess it was written by a few guys who were after their PhD.
As I mentioned earlier, matching your antenna reasonably well to 50 ohms is usually a matter of an L plus a C for the 2.4GHz band, $0.10 of parts. If efficiency is important to you, your design may benefit from wirewound inductors which are less lossy than multilayer, but they're also higher Q, so you may find you suffer reduced bandwidth in your matching network. Standard C0G/NP0 ceramics are usually fine for the caps. There are a few online impedance matching calculators designed just for the job of figuring out the LC network. You enter your centre frequency with source and load complex impedances and they figure out the values for you. Usually there are two options, one high pass and the other low pass. Which one you choose is up to you, but as I mentioned you might end up with one that, when added to your balun filter, ends up with two Cs or Ls either in parallel or series, and so you can merger them into a single part. Equally, sometimes, one of the options might present a part that is simply silly, like 0.01pF or 100pH, in which case choose the other option.