My biggest frustration with my car's built-in navigation is that when I ask it for the "nearest" item, more than 50% of the time it tells me to make a U-turn because it picked the mathematically nearest item, rather than picking the conventionally nearest item.
There are all sorts of ways to handle this request, but Siri's is the laziest, and in doing so asks more of the user's attention. If you just say "McDonalds" to either voice assistant, you also get a list. From that perspective, it's as if Siri ignores any information that might be gleaned from the rest of your sentence.
Siri learns from us but we also learn from Siri, and we might find that some words don't matter. You mind as well just say "McDonalds" if Siri is going to ignore the rest.
Reminds me of some of the chatbots that I use to make. No matter how hard I tried to test them, they always worked as expected. The minute someone else used it, it fell apart. Subconsciously, I knew what would work and what wouldn’t.
In my case, there were 9 McDonalds within a 10 mile radius.
In cases where there was only one location nearby, it took me right to it.
It's very hard for us to discuss Siri's internal state; by contrast it's easier to discuss Google's observable performance, which is to semantically differentiate between these two requests.
You can just ask Google, "Nearby McDonalds" and you'll get a list. "Go to the nearest McDonalds" and you get navigation.
“McDonalds” - gives me a list.
“Take me to the closest McDonalds” - brings up maps and starts navigating to the closest McDonalds.
After further experimentation. Siri doesn’t understand Nearest but does understand “closest” to mean that I don’t want a list.