I just got my DNA test back and it says I'm 8% Italian. None of my known ancestors are Italian. What's going on?
Welcome to the club. This happens to a lot of us.
A few possibilities, in rough order of likelihood:
One - the test is wrong in the most boring way. Ancestry estimates have wide confidence intervals, and an "8% Italian" result might be saying "we found a pattern that could be Italian, or could be some other Southern European population, with moderate confidence." Look at the underlying confidence range if the company shows it. Often the same result reads as "0–15% Italian" if you click into the details.
Two - your family tree is incomplete further back than you realise. Most people can name their great-grandparents. Beyond that, things get hazy. An Italian ancestor seven generations ago would contribute roughly 0.8% to your DNA - not exactly 8%, but in the right neighbourhood. Migrations, surname changes, and adoptions further back than three generations are surprisingly common.
Three - the reference panels are imperfect. Some European populations have overlapping DNA signals. Italian, Greek, and Balkan ancestry can look similar to the algorithm. The 8% might be Greek, Balkan, or even broadly Mediterranean.
If you really want to investigate, upload your raw DNA to a different company. If both call you 8% Italian, the signal is probably real. If only one does, it's a quirk of that company's algorithm.
Sunita's answer is solid. Adding one more point.
Reference populations are built from people who self-identify as having ancestry from a specific region for several generations. But "Italian" is a relatively recent national identity - for most of European history, what's now Italy was a collection of city-states, kingdoms, and shifting populations with significant gene flow from North Africa, the Middle East, the Balkans, and Northern Europe.
What the algorithm calls "Italian" is really "the genetic signature of people whose recent ancestors lived in the geographic area now called Italy." That signature overlaps with several neighbouring signatures. An 8% reading could legitimately be picking up signal from anywhere in that overlap.
Take ancestry percentages as broad strokes, not precise truths.