Abstract Modelling the progress of geochemical processes in CO2 storage sites is frustrated by uncertainties in the rates of CO2 flow and dissolution, and in the rates and controlling mechanisms of fluid–mineral reactions that stabilise the CO2 in geological reservoirs. Dissolution of CO2 must be controlled by the complexities of 2-phase flow of CO2 and formation brines and the smaller-scale heterogeneities in the permeability in the reservoirs which increase the fluid contact areas. The subsequent fluid mineral reactions may increase storage security by precipitating CO2 in carbonate minerals but the consequences of fluid–mineral reactions on caprock rocks or potential leakage pathways up fault zones are less certain as the CO2-charged brines may either corrode minerals or decrease permeabilities by precipitating carbonates. Observations from CO2-injection experiments and natural analogues provide important constraints on the rates of CO2 and brine flow and on the progress of CO2 dissolution and mineral–fluid reactions. In these experiments brines in contact with the propagating plume appear to rapidly saturate with CO2. Dissolution of the CO2 drives the dissolution of oxide and carbonate minerals, on times scales of days to weeks. These reactions buffer fluid pH and produce alkalinity such that carbonate dissolution moves to carbonate precipitation over time-scales of weeks to months. The dissolution of Fe-oxide grain coatings and the release of Fe to solution is important in stabilising insoluble Fe–Mg–Ca carbonate minerals but the rate limiting step for carbonate mineral precipitation is the transport of CO2-charged brines and silicate mineral dissolution rates. Observations from CO2-EOR experiments and natural analogues suggest that the silicate mineral dissolution reactions are initially fast in the low pH fluids surrounding the CO2 plume but that reaction progress over months to years drives minerals towards thermodynamic equilibrium and dissolution rates slow over 2–5 orders of magnitude as equilibrium is approached. The sluggish dissolution of silicate minerals is likely to preside over the long-term fate of the CO2 in geological reservoirs. Observations from injection experiments and natural analogues suggest that the potentially harmful trace elements mobilised by the drop in pH are immobilised as adsorbed and precipitated phases as fluid pH is buffered across mineral reaction fronts. There are very few observations of caprock exposed to CO2-rich brines. Preliminary examination of core recently recovered from scientific drilling of a natural CO2 accumulation in Utah suggests that the diffusion of CO2 into reservoir caprocks drives dissolution of Fe-oxides but subsequent precipitation of carbonate minerals likely retards the diffusion distance of the CO2. At this site thin siltstone layers are shown to be effective seals to the CO2-charged fluids, which has significant implications for the long term security of CO2 in geological reservoirs.