Mould growth is indicative of unhealthy indoor environments and, with a warming climate, increasingly poses a health risk. Understanding the prevalence and scope of the exposure largely relies on resident self-diagnosis; yet there is little guidance on how to optimise self-reported measures of mould in homes to achieve more accurate diagnosis of exposure. We compared the predictive performance of a range of self-reported measures that varied by their vernacular, framing, reference period, and severity of mould to be identified, against measures of mould taken from dust samples in 100 homes and analyzed using the quantitative polymerase chain reaction (qPCR) tests. Kappa and areas under the receiver operating characteristic curve (AUC) statistics were used to test the validity and accuracy of self-diagnosis of domestic mould. We find moderate agreement between self-reported and lab tested mould measures. Occupants tended to overestimate the presence of mould when asked about visible mould and suspicion of mould and to underestimate the presence of mould when asked about mould size, odour, dampness, and water damage. Identification of visible mould had the highest sensitivity while identification of mould larger than an A4 sheet of paper had the highest specificity. Combining self-reported visible mould and mould size achieved the best accuracy. When using self-rated mould severity (no, mild, moderate, or severe mould), grouping mild, moderate, and severe mould best detected actual mould presence. Prediction accuracy also varies by occupant sociodemographic and residential factors, with older age, lower household income, and major plumbing problems associated with better accuracy of self-diagnosed mould.