The prospects of using abundance ratios as stellar age indicators appear promising for solar analogues, but the usefulness of this technique for stars spanning a much wider parameter space remains to be established. We present abundances of 21 elements in a sample of 13 bright FG dwarfs drawn from the Kepler LEGACY sample to examine the applicability of the abundance-age relations to stars with properties strongly departing from solar. These stars have precise asteroseismic ages that can be compared to the abundance-based estimates. We analyse the well-known binary 16 Cyg AB for validation purposes and confirm the existence of a slight metal enhancement (∼0.02 dex) in the primary, which might arise from planetary formation and/or ingestion. We draw attention to systematic errors in some widely used catalogues of non-seismic parameters that may significantly bias asteroseismic inferences. In particular, we find evidence that the ASPCAP T eff scale used for the APOKASC catalogue is too cool for dwarfs and that the [Fe/H] values are underestimated by ∼0.1 dex. In addition, a new seismic analysis of the early F-type star KIC 9965715 relying on our spectroscopic constraints shows that the star is more massive and younger than previously thought. We compare seismic ages to those inferred from empirical abundance-age relations based on ages from PARSEC isochrones and abundances obtained in the framework of the HARPS-GTO programme. These calibrations depend on the stellar effective temperature, metallicity, and/or mass. We find that the seismic and abundance-based ages differ on average by 1.5–2 Gyr, while taking into account a dependency on one or two stellar parameters in the calibrations leads to a global improvement of up to ∼0.5 Gyr. However, even in that case we find that seismic ages are systematically larger by ∼0.7 Gyr. We argue that it may be ascribed to a variety of causes including the presence of small zero-point offsets between our abundances and those used to construct the calibrations or to the choice of the set of theoretical isochrones. The conclusions above are supported by the analysis of literature data for a larger number of Kepler targets. For this extended sample, we find that incorporating a T eff dependency largely corrects for the fact that the abundance-based ages are lower/larger with respect to the seismic estimates for the cooler/hotter stars. Although investigating age dating methods relying on abundance data is worth pursuing, we conclude that further work is needed to improve both their precision and accuracy for stars that are not solar analogues.