paternity tests show much higher incidence of non paternity, presumably because the self-selected population took the test because they had some doubt.
On the other hand, those with doubts are less-likely to have paternal information included in other genetic studies because they don’t want to know. So those studies show lower rates than what reality probably is.
Takes a lot of meta analysis, but best guess is probably 3-4% for all Americans.
Varies greatly by other demographic factors.