Abstract.  The fraction of urban volatile organic compound (VOC) emissions attributable to fossil fuel combustion has been declining in many parts of the world, resulting in a need to better constrain other anthropogenic sources of these emissions. During the National Institute of Environmental Research (NIER) and National Aeronautics and Space Administration (NASA) Korea-United States Air Quality (KORUS-AQ) field study in Seoul, South Korea, during May–June 2016, air quality models underestimated ozone, formaldehyde, and peroxyacetyl nitrate (PAN), indicating an underestimate of VOCs in the emissions inventory. Here, we use aircraft observations interpreted with the GEOS-Chem chemical transport model (version 13.4.0) to assess the need for increases in VOC emissions and for a revised chemical mechanism to improve treatment of VOC speciation and chemistry. We find that the largest needed VOC emissions increases are attributable to compounds associated with volatile chemical products, liquefied petroleum gas (LPG) and natural gas emissions, and long-range transport. Revising model chemistry to better match observed VOC speciation together with increasing model emissions of underestimated VOC species increased calculated OH reactivity by +2 s⁻¹ and ozone production by +2 ppb h⁻¹. Ozone increased by +6 ppb below 2 km and +9 ppb at the surface, and formaldehyde and acetaldehyde increased by +30 % and +120 % aloft, respectively, all in better agreement with observations. The larger increase in acetaldehyde was attributed to ethanol emissions, which we found to be as important for ozone production as isoprene or alkenes. The increased acetaldehyde significantly reduced the model PAN bias. The need for additional unmeasured VOCs, however, was indicated by a remaining model bias of −0.8 ppb in formaldehyde and a −57 % and −52 % underestimate in higher peroxynitrates (PNs) and alkyl nitrates (ANs), respectively. We added additional chemistry to the model to represent an additional six PNs from observed VOCs but were unable to account for the majority of missing PNs. However, four of these PNs were modeled at concentrations similar to other commonly measured PNs (> 2 % of PAN) indicating that these should be measured in future campaigns and considered from other VOC emission sources (e.g., fires). We hypothesize that emissions of oxygenated VOCs (OVOCs) such as ≥ C5 aldehydes from cooking and/or alkenes associated with volatile chemical products could produce both PNs and ANs and improve remaining model biases. Emerging research on the emissions and chemistry of these species will soon allow for modeling of their impact on local and regional photochemistry.