An ethnicity estimate is the percentage breakdown of geographic or ethnic regions that a DNA testing company generates from your autosomal DNA. The pie chart looks precise, but every number in it is a statistical probability produced by comparing your DNA to a reference panel of other people, not a direct measurement of where your ancestors lived. The same saliva sample, tested by 2 different companies, can come back with different percentages, and the same kit can change after the company's next update.
How the Percentage Gets Calculated
Every estimate works the same basic way: a company compares segments of your DNA to a reference panel of people with long, documented roots in one place, and assigns each segment to whichever population it resembles most. 23andMe brought a version of this to consumers in March 2008, sorting DNA into just 3 categories: African, European, and Asian, then replaced that tool with the more detailed Ancestry Composition in December 2012. AncestryDNA launched its own autosomal test, with ethnicity estimates included, in May 2012, then issued its first major revision in September 2013; MyHeritage did not add a comparable estimate until June 2017. Each company built, and still owns, its own reference panel and algorithm; none of it comes from a government record.
Why the Number Is a Probability, Not a Measurement
AncestryDNA tests its own reference panel with a leave-one-out analysis: it removes 5 percent of the panel, predicts the origin of the people removed using the other 95 percent, and repeats this 20 times until everyone has been tested. This correctly identifies a person's own reference region an average of 78.9 percent of the time, but accuracy varies widely. The Japan region was identified correctly close to 100 percent of the time. The Basque region was correct only 54 percent of the time, because Basque DNA closely resembles DNA from neighboring Spain and France. 23andMe calls its default ethnicity view the Speculative Estimate, which requires only 50 percent confidence before assigning a region, a level at which an assignment is about as likely to be right as wrong.
Continental Accuracy Versus Country-Level Accuracy
These estimates work best at a broad, continental level, telling European ancestry apart from East Asian, African, or Indigenous American ancestry with reasonable confidence. They work far less well at distinguishing between neighboring countries or regions within the same continent. The Basque example above is typical: populations that have shared centuries of migration and intermarriage across a border often carry DNA that is difficult for any algorithm to tell apart, which is why a specific country-level breakdown deserves less confidence than the continental grouping it sits inside.
Why Results Change Over Time and Between Companies
AncestryDNA has revised its reference panel repeatedly since 2013; its 2023 update alone took the region count from 84 to 88. An October 2024 overhaul renamed the feature Origins and added 24 new regions, 54 new subregions, and, for the first time, 64 ethnic groups within Africa. 23andMe added new African and East Asian detail in an August 2018 update, an acknowledgment that its reference panel had been thinner for those populations. MyHeritage nearly doubled the number of ethnicities it distinguishes, from 42 to 79, in a February 2025 revision. Because every company keeps a different reference panel, a different algorithm, and often an unstated time period, the same raw DNA file can be read into different regions at different companies. Even siblings can receive different percentages from each other, sometimes from a genuinely different share of a grandparent's DNA and sometimes from the reference panel itself.
Using the Estimate With Your Family Tree
An ethnicity estimate also reflects ancient population history more than recent family history. The regions on the pie chart describe where a population's DNA has been common over centuries, and sometimes thousands of years, of migration and settlement, not necessarily where a specific great-grandparent was born. A region can appear in your results because of population movements long before any documented ancestor, so it will not always correspond to a place your family actually lived within the 2 or 3 centuries that written records cover.
Treat a large regional percentage, 20 percent or more, as reasonably dependable, and a small trace amount, under about 5 percent, as the part most likely to shift at the next update. An ethnicity estimate cannot name a specific ancestor, confirm a family story, or substitute for documentary research. Sometimes a paper trail and a list of DNA matches will tell you more about a family's origin than the percentage ever will. Keep in mind that the estimate is a statistical opinion about your DNA today, not a fixed fact about your family.
Need help finding more records? We have genealogy research services available. You can also try our genealogical records directory which has more than 1.3 million sources to help you more easily locate the available records.