Climate trends on timescales of 10s to 100s of millions of years are controlled by changes in solar luminosity, continent distribution, and atmosphere composition. Plate tectonics affect geography, but also atmosphere composition through volcanic degassing of CO2 at subduction zones and midocean ridges. So far, such degassing estimates were based on reconstructions of ocean floor production for the last 150 My and indirectly, through sea level inversion before 150 My. Here we quantitatively estimate CO2 degassing by reconstructing lithosphere subduction evolution, using recent advances in combining global plate reconstructions and present-day structure of the mantle. First, we estimate that since the Triassic (250-200 My) until the present, the total paleosubduction-zone length reached up to ∼200% of the present-day value. Comparing our subduction-zone lengths with previously reconstructed ocean-crust production rates over the past 140 My suggests average global subduction rates have been constant, ∼6 cm/y: Higher ocean-crust production is associated with longer total subduction length. We compute a strontium isotope record based on subduction-zone length, which agrees well with geological records supporting the validity of our approach: The total subduction-zone length is proportional to the summed arc and ridge volcanic CO2 production and thereby to global volcanic degassing at plate boundaries. We therefore use our degassing curve as input for the GEOCARBSULF model to estimate atmospheric CO2 levels since the Triassic. Our calculated CO2 levels for the mid Mesozoic differ from previous modeling results and are more consistent with available proxy data.