A central challenge for sustaining international cooperation to cut global greenhouse gas emissions is confidence that national policy efforts are leading to a meaningful impact on the climate. Here, we apply a detection protocol to determine when the measurable signal of atmospheric CO2 can be distinguished from the noise of the carbon cycle and uncertainties in emission trends. We test that protocol with a database of 226 emission mitigation scenarios—the universe of scenarios vetted by the Intergovernmental Panel on Climate Change. These scenarios are descriptive of ‘baseline’ trajectories of emissions trends in the absence of new policies along with trajectories that reflect substantial policy efforts to stop warming at 1.5 °C–2 °C above pre-industrial levels, as embodied in the Paris Agreement. The most aggressive mitigation scenarios (i.e. 1.5 °C) require 11–16 years to detect a signal of demonstrable progress from the noise; 2 °C scenarios lengthen detection by at least a decade. As more climate policy analysts face the reality that goals of 1.5 °C–2 °C seem infeasible, they have developed ‘overshoot’ scenarios with emissions that rise above the agreed goal and then, later on, fall aggressively to achieve it. These pathways come at the political cost of a 1–2 decade delay in detection, even for the 1.5 °C scenarios. The Paris Agreement requires a global ‘stocktake’ that interrogates national mitigation efforts; our results suggest that this effort must grapple with the question of when the world can gain confidence that the diplomacy on climate is demonstrably making an impact.