Although there are certain rules about primer lengths, sometimes with AT rich regions it can be necessary to deviate from the suggested 18-29 nucleotide length and move toward 40 nt in length. Since GC pairs have three bonds it can also be helpful to pick primers that have G and C at thCertain e ends forming a "GC clamp". You should design a few forward and reverse primers just in case and run them on a thermocycler with a variable heat setting to test a gradient of temperatures from 46 to 62 degrees C. One will most likely work. Some PCR master mixes also help to reduce non specific binding by stabilization of the DNA- Qiagen makes a 2X ready to use master mix that helps with cloning of AT rich regions. 

You can also try to add 1 uL (25 uL in total) or 2 uL (50 uL in total) of DMSO in to the PCR mix, which helps to reduce non-specific binding.

It might be because of low GC and you should design a longer primer and also another thing to do is gradient of annealing tempratures using and also TouchDown PCR for non specific reducing.

if you mean point mutations in your desired sequence, I don't think longer primers will help. Rather change conditions, did you use proof-reading polymerase? Additives rather tend to decrease the specificity of amplicon, so try to avoid them.

Trying  other primers is not the first choice. Testing all possible PCR conditions with your present primers using a PCR machine with gradient of annealing temperatures is mandatory. If it does not work try other primers localized in other locations in order to amplify a longer fragment with a GC content which might work between 45°C to 55 °C

Adding DMSO is usually used with rich GC content of the fragment to amplify .

The first anwer from Thomas Weppelmann · is quite complete .

Using proof-reading TaqPolymerase is also a good point