That's pretty good [bad]. I'd really like to see a thorough, rigorous test of all the assistants. I haven't found one yet that doesn't have severe limitations.
We need a test that goes something like this: Queries about 1) trivia A, B, C, D... (sample of n >= 60), 2) playing A, B, C, D... songs or playlists (n >= 60), 3) directions to A, B, C, D... (n>=60), and so on.
The tests all need to be done in the best conditions (e.g., Siri on iPhone X, Google Assistant on Pixel 2, Alexa on Fire HD 10 in a sound-controlled room) by one person. Then, repeat the tests with 30 or more other people of varying accents and enunciations. Then repeat the all the tests in 'normal' environments.
*If someone does this and gets it published I'd like at least an acknowledgement. 😉