> for something like VOCs, it's likely using a CO₂ sensor and a lookup table that's static and only really valid for the test room it was created in.

It's the other way around, usually. CO₂ is more typically the estimated value, since its direct detection involves a relatively expensive NDIR value. VOCs as a category need some sort of modeling and estimation, but sensors can go fairly far by directly measuring a few species (hydrogen, for example), and hoping.