A CD burner works by firing precise beams of IR or UV laser radiation at a targeted point on the disc surface, where it interacts with a chemical dye. This dye changes optical reflectivity when exposed to such radiation, hence the term "burning". When reading it back, a lower intensity beam (usually at a different electromagnetic wavelength) scans over the disc and interprets the levels of reflectivity as a signal. When appropriately interpreted, this produces a stream of 0s and 1s.
There are two primary ways of encoding such data. The first is simply to assign "reflective" and "non-reflective" as 0 or 1 directly. One downside with this is that it is prone to errors. A speck of dust or a small scratch could corrupt the data, so extensive error detection and correction (e.g. parity) data blocks must be included. An alternative method is to use transitional encoding, which converts the transition of reflective to non-reflective and vice versa as 0 and 1 bits. This helps improve error detection and correction because long runs of reflective or non-reflective responses are simply ignored. Parity is still required, but to a lesser extent.
If the former method is used, it may be possible to etch more non-reflective bits (whether these are determined as 0 or 1, I am unsure). However, in my understanding, this simple version of encoding isn't commonly used.
If the latter method is used, things get more complicated. Causing a reflective bit to become non-reflective would break the transition (i.e. you'd have non-reflective followed by non-reflective) and simply destroy the bit. The parity would then correct the error, assuming there is enough remaining data. Since the read controller is aware of its position on the cylinder, trying to damage a single bit-pair in the hopes of offsetting the data in a way that allowed you to control it wouldn't work, as it would recognise that the bit-pair was destroyed and move on to the next bit pair, rather than shifting the stream by one bit.
The only trick I could see being feasible is to destroy the data to the point where the parity correction results in the wrong data, but this would not allow you to arbitrarily alter data - the results would be bound to a strict set of outcomes, based on the calculation of the parity. As such, it is relatively infeasible to modify the data on the disc in any meaningful way.