Carbon dioxide capture is essential to achieve net-zero emissions. A hurdle to the design of improved capture materials is the lack of adequate tools to characterise how CO2 adsorbs. Solid-state nuclear magnetic resonance (NMR) spectroscopy is a promising probe of CO2 capture, but it remains challenging to distinguish different adsorption products. Here we perform a comprehensive computational investigation of 22 amine-functionalised metal-organic frameworks and discover that 17O NMR is a powerful probe of CO2 capture chemistry that provides excellent differentiation of ammonium carbamate and carbamic acid species. The computational findings are supported by 17O NMR experiments on a series of CO2-loaded frameworks that clearly identify ammonium carbamate chain formation and provide evidence for a mixed carbamic acid - ammonium carbamate adsorption mode. We further find that carbamic acid formation is more prevalent in this materials class than previously believed. Finally, we show that our methods are readily applicable to other adsorbents, and find support for ammonium carbamate formation in amine-grafted silicas. Our work paves the way for investigations of carbon capture chemistry that can enable materials design.