Three poly(arylene ether ketone)s (PAEKs) with propylidene (C1, C2) and phtalide (C3) fragments and one phtalide-containing polyarylene (C4) were synthesized. Their chemical structures were confirmed via 1H NMR, 13C NMR and 19F NMR spectroscopy. The polymers have shown high glass transition temperature (> 155 °C), excellent film-forming properties and high free volume for this polymer type. The influence of various functional groups in the structure of PAEKs on gas transport parameters was evaluated. Expectedly, due to higher free volume the introduction of hexafluoropropylidene group to PAEK resulted in higher increase of gas permeability in compar-ison with propylidene group. The substitution of the fluorine-containing group on a rigid phtalide moiety (C3) significantly increases glass transition temperature of the polymer while gas permea-tion slightly decreases. Finally, the removal of two ether groups from PAEK structure (C4) leads to a rigid polymer chain that is characterized by highest free volume, gas permeability and diffu-sion coefficients among the PAEKs under investigation. Also, theoretically predicted transport parameters were investigated to further study the structure-properties relationship for the PAEKs. Methods of modified atomic (MAC) and bond (BC) contributions were applied for this purpose (estimation of gas permeation and diffusion). Gas solubility coefficients for PAEKs were fore-casted by “Short polymer chain surface based prediction” (SPCSBP) method. Both MAC and BC techniques showed reasonable predicted parameters for three polymers while a significant un-derestimation of gas transport parameters was observed for C4. Results for all three prediction methods were compared with the experimental data obtained in this work. Predicted parameters were in good agreement with experimental data for phtalide-containing polymers (C3 and C4) while for propylidene-containing poly(arylene ether ketone)s they were overestimated due to a possible influence of propylidene fragment on indices of oligomeric chains. MAC and BC methods demonstrated better prediction power than SPCSBP method.