> They aren't truly geographical- they're based on population structure in the genome.

Determining whether people are genetically similar is indeed well-studied. The problem is saying with certainty that you know where their ancestors lived.

There are a few ways the results end up being inaccurate, and they apply to all the companies.

The first and most important is that the data is from where people live now or in the very recent past (as I alluded earlier). That means they can only tell you how closely-related you are to people in a certain geographic area, whether those people have been there for 100 years or 1,000 years.

The second is that the data sets are filled with mostly European data, so anyone with even a bit of non-European ancestry is going to have less-accurate results[1] presented to them with a misleading degree of confidence.

The third is that there isn't an objectively "correct" algorithm for determining if someone is from one place or another. Every company has their own, which is why people get (sometimes wildly) different results from different companies.[2][3]

And the fourth is that the data collection doesn't seem to be especially accurate, perhaps related to the quantity or variety of SNPs that the testing company collects data about.

> Above you said "ex has a doctorate in genetics".

Correct. And the same way I take climate scientists' word for it that climate change is happening, instead of getting 10 degrees in different fields and gathering data myself, I understood her explanation and trust her expertise. Being a science consumer is always about trusting other people who have done the research, not about learning it all yourself and gathering your own data.

If you can find someone with a doctorate in genetics who has no affiliation with testing companies and believes these tests accurately tell someone where their ancestors were from, I'd be happy to consider the counterpoint. I have never seen such a thing in my research.

Regardless, all of the commercial operations have given twins differing results, so we can be fairly certain that any accurate way of determining genealogy from DNA has not been commercialized.

These tests, however, are good for finding relatives or determining how closely-related two people are.

1. https://www.nih.gov/news-events/previous-genetic-association...

2. https://www.livescience.com/63997-dna-ancestry-test-results-...

3. https://www.vox.com/science-and-health/2019/1/28/18194560/an...

DNA testing company vanishes along with its customers' genetic data