They got in a no mans land.

A "normal" 16-bit processor can span 64kB of address space. The same as all 8-bit processors. So they get a slight performance edge when working with data sizes larger than 8 bits because of the slightly larger ALU. But the difference is so small that fast 8-bit processors wins over slow 16-bit processors.

When having external memory, 16 and 32-bit chips were hampered by the amount of pins needed for the memory interface, so the 8-bit chips managed to keep ahead or at least alive.

Now we have 32-bit processsors that do solve the age-old memory address problems. And since they have "enough" RAM and flash internally, they don't need to consume a lot of I/O for any memory interfaces.

So the 16-bitters did work well for PC-class processors where all pins were dedicated to power and a memory interface. That market have since many years been totally overrun by 32-bit and 64-bit processors.

And for the microcontroller market, the 16-bit chips don't bring anything to the table that makes them able to really take market shares from 8-bit and 32-bit chips. That a 16-bit core is trivially small would have been an advantage except that the 32-bit cores are also trivially small with new process technologies.

No power consumption advantage.
No speed advantage.
No die size advantage.
No address space advantage.

The only advantage they may have is that a cheap microcontroller clone based on the 80186 or 80286 instruction set (but with 486-class instruction speeds) could make use of all the free 16-bit PC development software. Alas, there are already lots of free 32-bit development software too so that single advantage is so very small that it doesn't really count anymore.