The photospheres of low-mass red giants show CNO isotopic abundances that are not satisfactorily accounted for by canonical stellar models. The same is true for the measurements of these isotopes and of the $^{26}$Al/$^{27}$Al ratio in presolar grains of circumstellar origin. Non-convective mixing, occurring during both Red Giant Branch (RGB) and Asymptotic Giant Branch (AGB) stages is the explanation commonly invoked to account for the above evidence. Recently, the need for such mixing phenomena on the AGB was questioned, and chemical anomalies usually attributed to them were suggested to be formed in earlier phases. We have therefore re-calculated extra-mixing effects in low mass stars for both the RGB and AGB stages, in order to verify the above claims. Our results contradict them; we actually confirm that slow transport below the convective envelope occurs also on the AGB. This is required primarily by the oxygen isotopic mix and the $^{26}$Al content of presolar oxide grains. Other pieces of evidence exist, in particular from the isotopic ratios of carbon stars of type N, or C(N), in the Galaxy and in the LMC, as well as of SiC grains of AGB origin. We further show that, when extra-mixing occurs in the RGB phases of population I stars above about 1.2 $M_{odot}$, this consumes $^3$He in the envelope, probably preventing the occurrence of thermohaline diffusion on the AGB. Therefore, we argue that other extra-mixing mechanisms should be active in those final evolutionary phases.