From you experience, which DFT method and basis sets are recommended for organic molecules optimization using Gaussian and provide a trusted energy and proprieties ( e.g UV-vis spectra)
Write me in case you encounter problems with the PubChem optimized geometry. In the Gaussian Support page you'll find basis-set details:
https://gaussian.com/basissets/
For full optimization with B3LYP, I recommend to use 6-311G basis set. In ascorbic acid you won't need polarization functions. This is a previous step.
Once you get a "stable geometry" you may proceed to a newer basis-set, lets say cc-pVTZ which gives you a quicker and more accurate solution than 6-311+G(d,p). cc-pTZV are consistent for correlation.
Choosing methods and basis for organic molecules depends upon the nature of the problem one is studying. You may find that for the DFT, basis set dependency is less for a particular functional. But if one is targeting the excited state ( may or may not use solvent), Then the choice of DFT functional becomes quite essential. Review articles are available where benchmarks are performed for various types of molecules for different excited properties; in your case, it could be the absorption spectrum.