Recent advances in medicine and life-expectancy gains have fueled multidisciplinary research into the limits of human lifespan [1-3]. Ultimately, how long humans can live for may depend on selection favoring extended longevity in our evolutionary past . Human females have an unusually extended post-reproductive lifespan, which has been explained by the fitness benefits provided from helping to raise grandchildren following menopause [5, 6]. However, formal tests of whether such grandmothering benefits wane with grandmother age and explain the observed length of post-reproductive lifespan are missing. This is critical for understanding prevailing selection pressures on longevity but to date has been overlooked as a possible mechanism driving the evolution of lifespan. Here, we use extensive data from pre-industrial humans to show that fitness gains from grandmothering are dependent on grandmother age, affecting selection on the length of post-reproductive lifespan. We find both opportunities and ability to help grandchildren declined with age, while the hazard of death of women increased greatly in their late 60s and 70s compared to menopausal ages, together implying waning selection on subsequent longevity. The presence of maternal grandmothers aged 50-75 increased grandchild survival after weaning, confirming the fitness advantage of post-reproductive lifespan. However, co-residence with paternal grandmothers aged 75+ was detrimental to grandchild survival, with those grandmothers close to death and presumably in poorer health particularly associated with lower grandchild survival. The age limitations of gaining inclusive fitness from grandmothering suggests that grandmothering can select for post-reproductive longevity only up to a certain point.