Reciprocal hreflang
Each page has to point at its counterpart and be pointed back at. A one way declaration does not work. Where there is no counterpart, nothing should be declared at all, because claiming one that does not exist is a false statement.
URLs chosen per language
URLs are a choice rather than a translation. The term searched in Turkish and the one searched in English can differ even when they describe the same page. The structure has to be set at the start, because changing it later means redirect debt.
Where the language comes from
Reading language from the URL is more robust than reading it from a cookie or a browser setting. A search engine carries no cookie, so a language it cannot read from the URL is invisible to it.
A policy for missing translations
What happens when a page has no counterpart yet has to be decided up front. Publishing a half translated page tends to leave it weak in both languages.