Translating the text is the visible part. The invisible part is where the imprint, privacy notice and cookie consent are fed from, which language lives at which URL, and how a search engine reads that.
Imprint and privacy notice name a specific legal entity. Copied into three language versions, they all have to change on the day the company details change, and in practice one gets forgotten. The correct setup holds those details in one source that every language reads from, so an address change is made once and shows up correctly everywhere.
The checkbox itself is easy; what is hard is having the consent mean something in the system. If analytics consent and marketing consent are not kept apart, rejecting one either switches off the other or switches off nothing. On this site the two are separate and measurement only runs on the consent actually given; we apply the same split in the projects we build.
If three languages of the same content sit at one address, a search engine cannot know which to show. Each language needs its own URL, its own canonical and an hreflang set pointing at the others. This site works exactly that way: German at the unprefixed root, English and Turkish under their own prefixes, all from one codebase.
Usually add. Rebuilding only comes up when the existing structure never anticipated multiple languages and each one would live as a separate copy. We first check whether the current setup carries three things: separate URLs, single-sourced mandatory pages, and field-level formats.
We do not produce German content by machine translation. The source text can come from you and we write the German version together; for technical product copy that is usually the fastest route. The content of legal texts belongs to partner law firms; we build the infrastructure that carries them.
Send the address. We can usually list what is missing on the German side in a first pass.