With mobile phone cameras approaching 20 MP while maintaining small sensor sizes, pixels dimensions are typically around 1 μm or lower. What about the lenses? Are they good enough to produce an Airy disk that can make the most out of the full sensor resolution?
You may easily check it if you photgraph an objective test chart - which will have converging lines for judging the resolution. If 2px or so wide lines look very pale on the photograph when using 100% zoom then you may be sure that the objective does not catch with resolution of sensor.
Hint: it does not. Even huge number of expensive objectives for 1,5x crop cameras (magnitude larger than smartphone camera sensor) does not catch up with 20 MP.
One possible reason for making so huge number of MP is that many additional details may be made perceptible with usage of strong sharpening - which will make the noise stronger though but still may be acceptible under bright lighting.
P.S. It is not possible to distinguish the objective blur from AA filter effect but the AA filter tends to be weak nowadays and ideally introduces blur with 1px radius, I do not expect the optical resolution to come even close to that.
Due to aberrations, realistic lenses do not perform at the diffraction limit, especially off axis. Going to smaller f-numbers further exasperates not only aberrations but also how rapidly images go out of focus due to the inverse square relationship between lens f-number and hyperfocal distance. The overall best-performing lenses have f-numbers around f/2.8. When you multiply the lens' real MTF by the pixels' real MTF to get the lens*pixel system MTF, you find that there is a point of diminishing returns on system resolution around 4x oversampling (when the optical spot size is 4x larger than the pixel pitch). Generally, for low f-number lenses (f/1.8 - f/2.8) that limits a useful pixel size to about 900 nm or larger. Furthermore, there is a big difference between the actual pixel count and the useful pixel count (given some lower boundary on usable MTF). At these resolution limits (e.g., around 900 nm) the total number of useful pixels is about 4 times less than the actual number of pixels. I will present the evidence for this in late October 2016 at the ISA conference.