An isotopic-independent, highly accurate potential energy surface (PES) has been determined for CO(2) by refining a purely ab initio PES with selected, purely experimentally determined rovibrational energy levels. The purely ab initio PES is denoted Ames-0, while the refined PES is denoted Ames-1. Detailed tests are performed to demonstrate the spectroscopic accuracy of the Ames-1 PES. It is shown that Ames-1 yields σ(rms) (root-mean-squares error) = 0.0156 cm(-1) for 6873 J = 0-117 (12)C(16)O(2) experimental energy levels, even though less than 500 (12)C(16)O(2) energy levels were included in the refinement procedure. It is also demonstrated that, without any additional refinement, Ames-1 yields very good agreement for isotopologues. Specifically, for the (12)C(16)O(2) and (13)C(16)O(2) isotopologues, spectroscopic constants G(v) computed from Ames-1 are within ±0.01 and 0.02 cm(-1) of reliable experimentally derived values, while for the (16)O(12)C(18)O, (16)O(12)C(17)O, (16)O(13)C(18)O, (16)O(13)C(17)O, (12)C(18)O(2), (17)O(12)C(18)O, (12)C(17)O(2), (13)C(18)O(2), (13)C(17)O(2), (17)O(13)C(18)O, and (14)C(16)O(2) isotopologues, the differences are between ±0.10 and 0.15 cm(-1). To our knowledge, this is the first time a polyatomic PES has been refined using such high J values, and this has led to new challenges in the refinement procedure. An initial high quality, purely ab initio dipole moment surface (DMS) is constructed and used to generate a 296 K line list. For most bands, experimental IR intensities are well reproduced for (12)C(16)O(2) using Ames-1 and the DMS. For more than 80% of the bands, the experimental intensities are reproduced with σ(rms)(ΔI) < 20% or σ(rms)(ΔI∕δ(obs)) < 5. A few exceptions are analyzed and discussed. Directions for future improvements are discussed, though it is concluded that the current Ames-1 and the DMS should be useful in analyzing and assigning high-resolution laboratory or astronomical spectra.