We studied the spread of influenza in the community between 1993 and 2009 using primary-care surveillance data to investigate if the onset of influenza was age-related. Virus detections [A(H3N2), B, A(H1N1)] and clinical incidence of influenza-like illness (ILI) in 12·3 million person-years in the long-running Royal College of General Practitioners-linked clinical-virological surveillance programme in England & Wales were examined. The number of days between symptom onset and the all-age peak ILI incidence were compared by age group for each influenza type/subtype. We found that virus detection and ILI incidence increase, peak and decrease were in unison. The mean interval between symptom onset to peak ILI incidence in virus detections (all ages) was: A(H3N2) 20·5 [95% confidence interval (CI) 19·7–21·6] days; B, 18·8 (95% CI 15·8·0–21·7) days; and A(H1N1) 17·0 (95% CI 15·6–18·4) days. Differences by age group were examined using the Kruskal–Wallis test. For A(H3N2) and A(H1N1) viruses the interval was similar in each age group. For influenza B there were highly significant differences by age group (P = 0·0001). Clinical incidence rates of ILI reported in the 8 weeks preceding the period of influenza virus activity were used to estimate a baseline incidence and threshold value (upper 95% CI of estimate) which was used as a marker of epidemic progress. Differences between the age groups in the week in which the threshold was reached were small and not localized to any age group. In conclusion we found no evidence to suggest that influenza A(H3N2) and A(H1N1) occurs in the community in one age group before another. For influenza B, virus detection was earlier in children aged 5–14 years than in persons aged ⩾25 years.