The stable carbon isotope composition of atmospheric CO₂ (&delta;¹³C-CO₂) is a critical tracer for distinguishing the contributions of different carbon sources and sinks. However, the atmospheric background CO₂ concentration is only approximately 420 ppm, which renders direct injection and measurement unachievable using conventional continuous-flow isotope ratio mass spectrometry and limits the widespread application of this technique in high-frequency monitoring scenarios. In this study, a gas preconcentration unit (PreCon) was coupled to an isotope ratio mass spectrometer (IRMS). Key parameters, including the trapping duration of the two-stage liquid nitrogen cold traps (T2/T3), sample injection volume, and sample vial pretreatment protocol, were systematically optimized, and methodological validation was conducted using multiple reference materials via multiple pretreatment pathways. The results demonstrated that the optimal trapping time for both T2 and T3 cold traps was 200 s, under which quantitative CO₂ trapping was achieved without detectable isotopic fractionation. The system background signal accounted for approximately 0.4 % of the signal intensity of a typical sample, exerting no significant interference on the measurement results. Over the injection volume range of 0.5&ndash;5.0 mL, a strong linear correlation was observed between injection volume and signal response (<em>R</em>² = 0.9998). The minimum effective injection volume was 0.5 mL (corresponding to approximately 8.9 nmol CO₂), with a replicate measurement precision of 0.02 &permil; for &delta;¹³C. Helium flush was identified as the optimal pretreatment approach for air samples stored in Labco vials. Measurements via manual and automatic injection showed excellent consistency, and the results agreed well with the certified values of gas matrix reference materials. The maximum deviation among the three pretreatment pathways (PreCon, GasBench, and dual inlet (DI)) was 0.11 &permil;, and the method enabled accurate determination of both solid carbonate and gas matrix reference materials. This proposed method achieves online high-precision &delta;¹³C measurement of atmospheric background CO₂ at the milliliter scale, and provides reliable technical support for atmospheric carbon cycle tracing and isotopic monitoring of urban carbon emissions.