I somewhat agree, but for me personally, "car without wheels" and "shirt without stripes" are about equal in terms of usage regularity.

I'd also say nobody uses the phrase "birds without flight", but instead "flightless birds".

I'd imagine people say "hairless dogs" a lot more often than "dogs without hair".

And the shirt examples show that there are many uses of the lead phrase "shirts without" that work just fine, but for some reason "stripes" really stands out far beyond the others. "Shirts without shoulders" is kind of a bizarre term to match, when almost all search results show "off the shoulder" or "cold shoulder" as the thing being matched.

There's a larger, nuanced point here: statistical ML based "AI" is simply an API over aggregated samples of human intelligence. The sampling suffers from bias (the "white men" example in this thread), gaps (the OP's example, which relies on "shirts without stripes" being an uncommon, though perfectly acceptable phrase), and many more such shortcomings of what is essentially "rote learning".