The 2018 Go user survey analysis should be released soon, but I'm guessing you're right that it's only used for a few things; however, those few things are pretty significant in terms of marketshare (distributed systems, backend applications, and general purpose tooling--this hypothesis could probably be checked against the StackOverflow survey). Not much on the front-end or mobile application side nor on the embedded side, but probably making some headway on the AI/ML/data-science side.
In general, I think of Go as Python but with 15 years of hindsight but also 15 years of market catchup. Go does a lot of things better than Python (performance, concurrency, distribution, deployment, etc), but the reverse isn't true. The principle advantage to Python is that it was the lingua franca for the data science niche before the niche became a mainstream competency--this means more people were exposed to Python rapidly, but it also means that people are beginning to build high-quality data science libraries in other languages, including Go, which closes the gap between Python and other languages.