There is pseudoscience, and then there is "pseudoscience". What I've found with the MBTI is that it's mostly just "pseudoscience", by the strict definition of empiricism, however, that doesn't make it completely useless pseudoscience. Let me explain:

It's a convenient mental framework. If someone fits one of the archetypes very strongly, then the framework will be very useful in predicting and recognizing how they think and behave in general. If someone doesn't really fit any of the archetypes very strongly, then yes it is useless, they'll get different results all of the time, and it will be pseudoscience. So while the test is bad at predicting even 60% of people's types correctly, people who do fit the archetypes will still find the test valuable.