At PyOPL index Julia is not 23th, but:
In France, Julia is 7th, the highest ever at 3.15 % (up 16 places or +2.5 %), after Scala (up 7), then JavaScript. Swift has a huge drop.
In the Germany, Julia is 12th (up 6 places), after Kotlin, also with huge gain, and just ahead of Swift.
In the US, Julia is 18th (up 6).
A July TIOBE index, Julia mentioned again:
Headline: C++ is about to surpass C
[…]
Elsewhere in the TIOBE index, we see that COBOL, after a long time, re-enters the top 20, while Julia is seriously approaching it.
So Julia is approaching COBOL, in goodness with 1.9…? Or just its rank. 
1- JavaScript / TypeScript - 915694 jobs
2 - Python - 603507 jobs
…
30 - Fortran - 252 jobs
31 - Julia - 96 jobs
It climbed 8 places in 1 year!
Any insights as to what were the defining factors for Julia climbing this year?
The primary one is likely that TIOBE is a garbage index where the methodology is made up and the numbers don’t matter.
We can agree on that. However, this can be a nice signal for those who actually believe the index is valuable and the top 20 languages are somehow worth picking up.
I can live with the idea of people “arriving in Julia community for the index, staying for the language”.
TIOBE gets some kind of broad strokes correct. It consistently ranks C and Python above Forth and Logo. Perl has trended strongly downward in the past 20 years, and Python strongly upward.
Click on the percents to see the history for the languages. The high frequency components are clearly noise. And it’s pretty noisy. For languages like Julia, their measurements show signal and noise of comparable size. (click to see, also, I don’t see a button for some languages, but you can edit the url by hand).
Between last November and last month, they show Python popularity dropped by 30%. We shouldn’t assign any significance to that.
Furthermore, their methodology has some built in biases that can be hard to measure.
For example in some languages its common to refer to “X scripting” in addition to “X programming”. The former does not count.
Or, “I’ve heard about sql queries, I’d like to learn what a query is and how to do it”. I find the same order of magnitude of hits for “sql programming” and “sql query”.
Lots of other things have been pointed out: For example, their time kernel for computing popularity is flat.
Another item which has probably been mentioned, but I don’t recall seeing it: Languages target, or have grown into, different markets. And markets have varying sizes.
They should call it “Whose Line of Code is it Anyway?” 
Maybe our SEO for “Julia programming language” via Documenter.jl finally started to kick in?
Julia searches and Tiobe ranking climbs around/after JuliaCon each year. It will settle again past #20 in the following months…