I asked my home’s voice assistant a simple question — where’s a family member right now — and got told, in effect, that it had never heard of them. Not “I don’t know where they are.” Flatly unaware they existed, despite being a real, fully set-up person in the system for years.
That answer was wrong in an interesting way — wrong enough that it was worth actually digging into rather than just giving up on the feature.
“I don’t know” and “I was never told” are different failures
A voice assistant failing to answer a question can mean a few different things, and they call for completely different fixes. It might genuinely not have the data — nobody’s tracking that person’s location at all. It might have the data but be confused by how the question was phrased. Or, the specific failure here, it might have the data sitting right there in the system, correctly tracked and totally usable, but never have been told that this particular assistant is even allowed to look at it.
That third case is invisible from the outside. The assistant doesn’t say “I’m not allowed to check that.” It just acts like the information doesn’t exist, because from its permitted point of view, it doesn’t.
Two separate bugs, stacked on top of each other
Digging in turned up two distinct problems, and both had to be fixed for the feature to actually work. First: most of the household’s location data existed in the system, but had only ever been explicitly shared with one voice integration (a different one, tied to a separate ecosystem) — never with the specific assistant being asked. A permissions gap, not a data gap.
Second, and more surprising: even once that was fixed for one person, asking by a nickname or a shortened first name still failed the exact same way. It turned out the system’s built-in question-answering only does exact matching against a name or a specifically registered nickname — no fuzzy matching, no “close enough.” A name spelled slightly differently, or a name close to correct but not registered, fails identically to a name it’s never heard at all. From the outside, both look like the same blank “I don’t know who that is” — but one is a real gap and the other is a one-word registration fix.
The actual fix took minutes once correctly diagnosed
Grant the right permission to the right assistant for each person’s location data, then explicitly register every nickname anyone in the household actually uses — not just legal first names. Once both were in place, the feature worked correctly for everyone in the house, using whatever short name or nickname people actually say out loud, not the name on their ID.
If a smart home feature “doesn’t work,” ask this first
- Is the underlying data actually being tracked somewhere in the system, or does it genuinely not exist yet?
- Has that data specifically been shared with the assistant or integration you’re asking, or only with a different one?
- Are you asking using the exact name the system has on file, or a nickname it’s never been told about?
A feature that “doesn’t work” is often a feature that was never actually asked correctly, or never given permission to answer. Worth the ten minutes it takes to check before assuming the technology just can’t do it.