<p>Research in machine recognition of handwritten text has predominantly focused on Latin scripts. There are a few works on Indic scripts. Of those, Gurmukhi script recognition is still at an embryonic stage, mainly due to the scarcity of data. The earlier studies are limited by incomplete character coverage, small datasets, and adult-only samples, ignoring age-related handwriting variability. In this work, we address these gaps by presenting the largest and most comprehensive Gurmukhi handwritten dataset to date, consisting of 335,245 Gurmukhi characters collected from 425 individuals aged 5 to 80 years. The novelty of this study lies in its complete basic character and numeral coverage, substantially larger dataset, and inclusion of handwriting samples from children, adults and old persons to analyse age-related handwriting variations. We employed a Convolutional Neural Network (CNN) architecture for recognition and evaluation. Our results demonstrated accuracy rates of 95.14% for children, 90.42% for old persons, and 97.40% for intermediate adults, suggesting the variations in handwriting. An overall accuracy of 93.14% for the mixed dataset was observed. To the best of our knowledge, this is the first work of this kind on the handwritten Gurmukhi dataset. The dataset has been deposited in Zenodo and is available with the permission of first author.</p>