Number of search results is an awful and arbitrary method for determining language popularity. For that matter, why not make a poll, ask an ouija board, or put a chicken in front of a keyboard to see what language it types (Perl would win that one)?
How about we look at actual code instead? Github, the most popular source code hosting service, regularly counts how much code is written by looking at pull requests. They publish the results here: https://octoverse.github.com/#...
When we look at actual code, the results are very different:
1) Javascript
2) Python
3) Java
4) Typescript
5) C#
6) PHP
7) C++
8) Shell
9) C
10) Ruby
Granted, plenty of companies use GitLab, Bitbucket, or use other version control systems internally, so GitHub does not account for all developers.
Still, I would argue it is a much better representation of the industry than search results.