Background: Adult obesity and insufficient physical activity co-evolve with economic development, urbanisation and digitalisation, which complicates interpretation of country-level associations. Objective: To examine whether within-country changes in insufficient physical activity were associated with adult obesity after accounting for internet penetration and other structural factors, whether internet penetration remained associated with obesity, and whether deviations from common GDP/year benchmarks co-varied. Material and methods: An unbalanced ecological panel included 2,707 country-year observations from 164 countries during 2000-2022. Two-way fixed-effects linear models with country-clustered standard errors estimated associations with adult obesity after adjustment for internet use, log GDP per capita, dietary energy supply, unemployment, urbanisation, population ageing, secondary enrolment and services employment. Obesity and inactivity benchmark residuals were defined as observed minus values predicted by log GDP per capita and year effects. Sensitivity analyses added health expenditure, excluded countries with fewer than five observations and decomposed the residual association into between-country and within-country components. Results: Insufficient physical activity was positively associated with obesity (β=0.174, SE=0.041, p<0.001); internet use also remained positively associated (β=0.037, SE=0.010, p<0.001). The inactivity benchmark residual was associated with the obesity benchmark residual in the fixed-effects model (β=0.163, SE=0.045, p<0.001), with similar estimates after health-expenditure adjustment (β=0.164) and restriction to countries with at least five observations (β=0.162). The unadjusted between-country and within-country components were both positive (β=0.285 and β=0.189, respectively; both p<0.001). Conclusions: Within countries, higher insufficient physical activity and higher internet use were associated with higher adult obesity. Country-years above the GDP/year benchmark for inactivity also tended to be above the corresponding obesity benchmark. These ecological results are associational; the residuals are model-dependent descriptive diagnostics, not validated indices or causal estimates.