Unintended local temperature enhancement by excitation laser might change Raman spectral features and potentially lead to misinterpretation of the data. To evaluate robustness of Raman CO2 densimeters in the presence of laser heating, we investigate the relation between temperature (T, °C), density (ρ, g/cm3), and Fermi diad split (Δ, cm−1) using a high‐pressure optical cell at 23°C to 200°C and 7.2–248.7 MPa. Results indicate that Δ decreases concomitantly with increasing temperature for a constant density in all density regions investigated. This result suggests that the density estimated based on Δ might be underestimated if the fluid is heated locally by the laser. Combining results of earlier studies with those of the present study indicates that the temperature dependence of Δ (|(∂Δ/∂T)ρ|) has a maximum value around 0.6–0.7 g/cm3. Consequently, at very high densities such as 1.1–1.2 g/cm3, |(∂Δ/∂T)ρ| is small. Thus, Δ at such densities is less affected by laser heating. However, at densities below approximately 0.7 g/cm3, although |(∂Δ/∂T)ρ| becomes smaller at lower densities, the relative density decrease becomes larger even for a small density decrease because the density itself becomes smaller. Therefore, at such densities, a density decrease of more than 10% was observed for some fluid inclusions, even at typical laser powers for inclusion analysis. Finally, to accurately estimate the density even in the presence of laser heating, we show that it is effective to estimate the intercept Δ from the correlation between Δ and laser power and substitute it into Δ–ρ relations.