Purpose It has been proposed that parents should be educated about child autistic spectrum disorder (ASD) 'red flag' traits to help professionals identify and address concerning behaviours as early as possible. This study aimed to empirically demonstrate that established/recognised 'red flag' traits in the first 3 years of life would reliably predict ASD risk severity in later childhood, associated with established ASD risk correlates and mirroring functioning diagnostic categories. Methods Using retrospective parental report data from the Mental Health of Children and Young People in Great Britain survey (N = 7977), latent class analysis (LCA) and a quasi-latent transition analysis were used to (1) identify profiles of variation in parent reports of child 'red flag' traits before and after age 3 and (2) model transitions in risk from 3 years and below to ≥ 3 years, respectively, per the 'optimal outcome' model. Results Three distinct classes, each characterised by variation in parent 'red flag' trait reporting were identified for the '≤ 3 years of age' and the '≥ 3 years of age' data. Both LCA class profiles comprised groups of children characterised by low, medium and high ASD risk. Dose-response effects for a number of recognised ASD correlates across the low, moderate and high risk '≥ 3 years of age' classes seemed to validate older classes in terms of ASD relevance. Over 54% of children characterised by the highest levels of ASD 'red flag' trait probability at 3 years and below (2% of sample), also populated the high-risk class evidenced in the '≥ 3 years of age' LCA. Conclusions Retrospective parental reports of child ASD 'red flag' traits ≤ 3 years of age were reliable indicators of ASD risk in later childhood.