Preamble:
GDP, is a bit of a synthetic metric.
As you point out, there's no purchase level data about what's imported vs. not.
The way this is handled is that this quarter's imports are set against this quarter's consumption - basically the method assumes the import/domestic mix of business inventories stays the same (true enough in the long run, very incorrect in short term shocks).
That's why extremely disingenuously the AP says:
>Trade deficits reduce GDP. But that’s just a matter of mathematics. GDP is supposed to count only what’s produced domestically, not stuff that comes in from abroad. So imports — which show up in the GDP report as consumer spending or business investment — have to be subtracted out to keep them from artificially inflating domestic production.
Answer:
What happened here[1] is that the BLS makes a bunch of assumptions to get data out in time (preliminary figures based on historical seasonal trends, etc.) but this quarter, their assumptions about consumer spend were far too aggressive.
It happens all the time, especially in strange times like 1Q was, but there's also career/political incentive to be aggressive on the advanced data, since that's what drives the big headlines.
[1]https://www.bea.gov/system/files/gdp1q25-3rd-chart-02.png