Background: Assessing health-related quality of life (HRQOL) in people with advanced dementia is challenging but important for informed decision-making. Proxy measurement of this construct is difficult and is often rated lower than self-report. Accurate proxy rating of quality of life in dementia is related to identification of concepts important to the person themselves, as well as the sensitivity of the measures used. The main aim of this study was to compare the performance of two instruments—QUALID and EQ-5D-5L—on measuring HRQOL in people with advanced dementia. Methods: In a sub-study nested within a cluster-RCT we collected proxy(nurse)-completed EQ-5D-5L and QUALID measures at baseline, 3, 6, 9 and 12 months’ follow-up for people with advanced dementia, residing in 20 nursing homes across Australia. Spearman’s rank correlations, partial correlations and linear regressions were used to assess the relationship between the HRQOL instrument scores and their changes over time. Results: The mean weight from 284 people for the EQ-5D-5L and QUALID at baseline were 0.004 (95% CI − 0.026, 0.033) and 24.98 (95% CI 24.13, 25.82), respectively. At 12 months’ follow-up, 115 participants remained alive. EQ-5D-5L weights and QUALID scores at baseline and at follow-up were moderately correlated (r = − 0.437; p < 0.001 at 12 months). Changes within QUALID and EQ-5D-5L across the same follow-up periods were also correlated (r = − 0.266; p = 0.005). The regression analyses support these findings. Conclusion: Whilst these quality of life instruments demonstrated moderate correlation, the EQ-5D-5L does not appear to capture all aspects of quality of life that are relevant to people with advanced dementia and we cannot recommend the use of this instrument for use within this population. The QUALID appears to be a more suitable instrument for measuring HRQOL in people with severe dementia, but is not preference-based, which limits its application in economic evaluations of dementia care.