Part 2: South Asian people getting flagged as Romani
Yeah, part 2 is here. After AncestryDNA wrongfully labeled me as 23% Eastern European Romani - while I believed I have Punjabi ancestry - and I have led countless arguments with prople who insisted I actually have Romani ancestry (as if there were no Indians in Europe, ever), had to explain I have extensive numbers of meaningful Hindu/Sikh matches and a consistent Punjabi/Haryanvi visual phenotype in my family, and lost my nerves a few times while doing so, I finally have a conclusion.
One of those people arguing with me sent me whitepapers written by AncestryDNA in years 2023 and 2025. They did it to support their own claim that Eastern European Roma is detected by AncestryDNA with 100% precision and therefor I must be Roma (yeah, because having 2nd and 3rd cousins from India and sharing 20-40 cM of DNA with them apparently means nothing at all according to those people).
2023:
https://archive.org/details/AncestryDNA-ethnicity2023whitepaper
2025:
https://archive.org/details/AncestryDNA-2025ancestralregionswhitepaper
Between those two revisions, I noticed AncestryDNA completely deleted any metrics of their confidence with Indian ancestry, and deleted their recall rates of Eastern European Roma. This observation further supports my theory that AncestryDNA has problem distinguishing between those two categories.
The root cause of what happened to me is (and this is only my hypothesis, that has very real foundations however):
Eastern European Roma (~95% precision, ~95% recall - based on the 2023 whitepaper) has ancestry originating between modern day Punjab, Rajasthan (possibly Haryana?), but AncestryDNA also attempts to distinguish them from modern day Northern Indians (but with shocking ~70% precision and ~70% recall - based on the 2023 whitepaper), which has a couple of problems:
- North India is too broad to confidently catch anything as specific as Punjab
- The claimed precision of 70% doesn’t help either
- They have only 925 samples for India, but India has 1.5 Billion people, and there is countless distinct groups within India
- Since Eastern European Roma have very geographically specific signal in their genome coming from Punjab or neighbouring locations (and precision and recall of ~95% - based on the 2023 whitepaper, and even 100% precision based on the 2025 whitepaper), AncestryDNA just confidently flags you as Eastern European Roma if you happen to have some of the same signature in your DNA - and I think that happens specifically because the category for modern Northern Indians is so immensely broad and uncertain in comparison to Eastern European Roma.
But since they also deleted any confidence metrics on their Indian ethnicity estimates, and at the same time the recall metrics for Eastern European Romani, it is reasonable to assume that AncestryDNA knew about the problem but decided not to disclose it.
TL;DR: AncestryDNA doesn’t publish their certainty with Indian ancestry anymore, which points to the possibility that the most accurate and confident reference they have for Punjab and neighbouring populations is Eastern European Roma, making it possible even for people born in India being classified as Eastern European Romani by AncestryDNA’s algorithms.
This didn’t happen only to me, but also 30+ of my matches (which are the ones I discovered so far in the ~10% of all of my matches that I manually checked).
Since majority of readers come from the US and Canada: Yes, we do have Indians in Europe. Yes, they are having kids. No, I am not 300 years old - Indians were certainly already here when my grandma was born - although in sparse numbers.
