I like to think of GraphQL as a specific tool to address a specific set of challenges. Is it convenient to be able to batch up a bunch of requests in one go for a frontend JS client? Would it be useful to be able to specify a subset of the fields to be returned in a query? If the answers to these questions and others are "yes," GraphQL might be a good fit for what one is doing.

I've personally had good experiences with it. It does bring in additional complexity. Is the complexity worth taking on in a specific dev team? This is something that will depend on the team.

It is usually the case that a tool is not useful outside of the set of challenges it is tailored to address. And in some cases a tool is not well designed or implemented. It's always good to use the right tool for the job, at the level of complexity that a team is in a good position to handle.