The composition and function of the human gut microbiota has been linked to health and disease. We previously identified correlations between habitual diet, microbiota composition gradients and health gradients in an unstratified cohort of 178 elderly subjects. To refine our understanding of diet–microbiota associations and differential taxon abundance, we adapted an iterative bi-clustering algorithm (iterative binary bclustering of gene sets (iBBiG)) and applied it to microbiota composition data from 732 faecal samples from 371 ELDERMET cohort subjects, including longitudinal samples. We thus identified distinctive microbiota configurations associated with ageing in both community and long-stay residential care elderly subjects. Mixed-taxa populations were identified that had clinically distinct associations. Microbiota temporal instability was observed in both community-dwelling and long-term care subjects, particularly in those with low initial microbiota diversity. However, the stability of the microbiota of subjects had little impact on the directional change of the microbiota as observed for long-stay subjects who display a gradual shift away from their initial microbiota. This was not observed in community-dwelling subjects. This directional change was associated with duration in long-stay. Changes in these bacterial populations represent the loss of the health-associated and youth-associated microbiota components and gain of an elderly associated microbiota. Interestingly, community-associated microbiota configurations were impacted more by the use of antibiotics than the microbiota of individuals in long-term care, as the community-associated microbiota showed more loss but also more recovery following antibiotic treatment. This improved definition of gut microbiota composition patterns in the elderly will better inform the design of dietary or antibiotic interventions targeting the gut microbiota.