Introduction Clinicians managing patients with acute recreational drug or new psychoactive substance toxicity typically depend on self-reported drug(s) used. This study compares patient self-report (and/or from other sources) to the substance(s) that were subsequently identified in serum. Methods A prospective sample of 1,000 adults presenting to a tertiary care, urban emergency department in London, United Kingdom, with acute recreational drug/new psychoactive substance toxicity was collected from 1 February 2019 to 2 February 2020. A total of 939 appropriate samples underwent qualitative analysis by high-resolution mass spectrometry with comparison to a database of drugs/metabolites. Data on the stated drug(s) used were extracted from the routine medical chart/records; results were batched by drug class, when appropriate, and analysis was performed using R software. Results Seven hundred and ninety-nine (85.1%) patients were male with a median (IQR) age of 34 years (27 to 42 years). Six hundred and thirty-five (67.6%) patients reported using two or more drugs. The median (IQR) positive predictive value of a self-report substance having been taken was 0.68 (IQR: 0.44–0.86); conversely, the median negative predictive value of a substance having not been taken was 0.90 (IQR: 0.53–0.95). There was variability in the accuracy of reporting. For example, self-reported opioid use had a 90.5% likelihood that opioids were detected on analysis, whereas hallucinogens were only detected in 18.8% of samples when use was reported. Individuals were also mostly accurate in not underreporting substances. For example, those not explicitly reporting gamma-hydroxybutyrate use were 97.5% truly negative. Discussion Overall, most users were relatively accurate in their self-report of what class of drugs they had used, although there was variability in this accuracy. However, other drugs were present even when not reported, for example, opioids with disproportionate detection of prescription and over-the-counter (non-prescription) opioids that were unreported. Conclusions Self-report (and/or collateral reports) had overall relatively high concordance with the likelihood that a substance was, or was not, recently used. Therefore, clinicians can make initial treatment decisions based on the self-reported drug(s) used in most cases.