Privacy

Is [XMPP's lack of popularity] really just a marketing/UX/UI problem?

No, there is more to it than that. Off the top of my head, these issues stand out as major hurdles:

• XMPP is a relatively basic protocol. By itself, it cannot compete because it lacks modern features like end-to-end encryption, persistent message history, group chat, etc. It does have extension specs (XEPs) to provide many such features, but it still lacks a single cohesive spec identifying and unifying the important ones. You could call it fragmentation. This makes it overly complex for implementors, and leads to the next problem:
• Someone wanting a messaging service with a competitive feature set must first identify at least one server that implements all the relevant XEPs, and a client for each of their devices that implements the same. That's not viable for most people, many of whom have only a vague notion of what a communications protocol is.
• Ever since public XMPP support was dropped by big services like Google and Facebook, the availability of reliable, free, public servers has all but vanished. Most people wouldn't know with confidence how to find one, let alone one with all the needed extensions. And even if they do find one, most will be unable to assess whether it will still be running in ten years or more. This makes it quite a gamble to tie your online identity and network of contacts to whatever server you find.

So, while XMPP (with appropriate extensions) is still a capable protocol, the expertise and support required to make it competitive is not readily available to most people. I might suggest it to small groups who have local expertise to get it all set up and keep things running well, but not to the general public.

Meanwhile, Matrix has a unified spec with a rich feature set, a variety of homeservers and client apps that support it, sufficient momentum for continued development, and the critical mass to make it viable for global public use.