No single DNA test is definitively the most accurate for ethnicity across all backgrounds. The best test for you depends on your specific ancestry. 23andMe generally offers the most granular worldwide coverage and a unique confidence-level tool that lets you see how reliable each part of your estimate really is. AncestryDNA has the largest customer database, which strengthens its ability to identify genetic communities. And Living DNA outperforms both for British and Irish ancestry specifically, breaking the region into 21 sub-regions no other company matches.
The deeper truth is that all ethnicity estimates are exactly that: estimates. Understanding how they work, where they excel, and where they fall short will help you interpret your results far better than simply picking the “right” company.
How Ethnicity Estimates Actually Work
Every DNA testing company compares your DNA to a reference panel, a curated database of people whose ancestry in a specific region has been verified, often going back several generations. The company’s algorithm scans hundreds of thousands of points across your chromosomes, looks for patterns that match those reference populations, and assigns percentages to each region.
The algorithms behind this process are sophisticated. They use statistical methods rooted in population genetics that recognize how chromosomes cluster into groups of similar patterns over short stretches of DNA. As the algorithm moves along a chromosome, the clustering can shift, reflecting the reality that different segments of your genome may trace back to different populations. This “chromosome painting” approach is why your results come with a colorful map of your chromosomes, each segment shaded by its estimated origin.
The critical limitation is the reference panel itself. If a company has 5,000 reference samples from Western Europe but only 200 from East Africa, its European estimates will naturally be more precise. This is why your background matters when choosing a test.
23andMe: Best for Worldwide Granularity
23andMe reports ancestry across more than 2,000 regions and populations, making it the broadest in global coverage. For people of African American descent, 23andMe scientists have identified 213 genetic communities of African descent within the United States alone, connecting customers to specific regional roots. That level of detail is unusual in the industry.
What genuinely sets 23andMe apart is its confidence slider, which lets you adjust your results between 50% and 90% confidence. At the default 50% setting, 23andMe calls its own results “speculative,” explicitly meaning they reflect conjecture or a good guess. At that level, you might see specific country-level breakdowns like 12% Italian or 8% Polish. Slide the confidence to 90%, which 23andMe labels “conservative,” and those country-level results largely disappear. In testing, country-specific signals in places like Italy, the British Isles, and Eastern Europe shrink to near zero, with roughly 75% of chromosomal segments reassigned to broad labels like “Broadly European” or “Broadly Northwestern European.”
This tells you something important: the specific country-level percentages that most people focus on are the least reliable part of any ethnicity estimate. The continental-level results (European, East Asian, Sub-Saharan African) are where the real accuracy lives. 23andMe is the only major company that makes this tradeoff visible to customers.
AncestryDNA: Best for Genetic Communities
AncestryDNA’s biggest advantage is its database size, currently the largest of any consumer test at over 25 million customers. More customers means more data points for the algorithm to work with, and it particularly strengthens Ancestry’s “Genetic Communities” feature, which groups you with people who share recent ancestors from the same town, county, or migration path.
For people with North American, British, Irish, or Northern European roots, AncestryDNA’s community assignments are often strikingly specific, sometimes pinpointing a particular county in Ireland or a migration route from Appalachia to the Midwest. These communities update roughly once a year, typically in the fall, as Ancestry incorporates new customers into its reference data and recalculates existing results. Your underlying DNA data doesn’t change, but the interpretation improves over time.
AncestryDNA does not offer a confidence slider like 23andMe, so you see only one version of your estimate. The percentages it displays are its best guess, without a built-in way to check how much of that guess is solid versus speculative.
Living DNA: Best for British and Irish Roots
If your primary interest is ancestry within the British Isles and Ireland, Living DNA is the standout choice. It divides the region into 21 distinct geographical sub-regions, including areas as specific as South Central England, South Wales border, and Devon. No other major company comes close to that resolution for Britain.
Living DNA builds these regional predictions by analyzing over 700,000 markers in your DNA and comparing them against a database of people with known, multi-generational ancestry in each area. The result is a level of specificity that can distinguish, say, someone with deep roots in Cornwall from someone with heritage in East Anglia. For people with ancestry outside the British Isles, Living DNA’s coverage is noticeably thinner, making it a specialist tool rather than a general-purpose test.
Why Results Differ Between Companies
It’s completely normal to get noticeably different ethnicity percentages from different companies. One might say you’re 40% English while another says 25%. This doesn’t mean your DNA changed or that one test is broken. Three factors drive these discrepancies.
- Reference panels differ. Each company curates its own set of reference populations, drawn from different sample sizes and sourced from different research projects. If Company A’s “French” reference group overlaps genetically with Company B’s “German” group, your DNA might be labeled French by one and German by the other.
- Regional boundaries are arbitrary. DNA doesn’t respect national borders. Populations in neighboring countries share enormous amounts of genetic material, and the lines a company draws between “Scandinavian” and “British” are statistical choices, not biological facts.
- Algorithms weight evidence differently. Some algorithms are more aggressive in assigning DNA to specific regions. Others default to broader categories when the signal is ambiguous. This is essentially what 23andMe’s confidence slider reveals: the difference between a bold guess and a cautious one.
Continental vs. Country-Level Accuracy
All major tests are highly accurate at the continental level. If you have significant European, East Asian, or Sub-Saharan African ancestry, every company will detect it reliably. The errors and disagreements almost entirely live at the country and sub-regional level.
This is because human genetic variation is genuinely organized by continent. Thousands of years of relative separation between continental populations created distinct genetic signatures that algorithms can identify with high confidence. Within a continent, though, the genetic differences between neighboring populations are far smaller and harder to distinguish. The difference between someone of Polish ancestry and someone of Ukrainian ancestry, genetically speaking, is subtle enough that current algorithms frequently confuse them.
If your family tree spans multiple continents, any major test will give you a broadly reliable picture of those proportions. If you’re trying to distinguish between two neighboring European countries, or pinpoint a specific region within West Africa, treat those percentages as educated guesses rather than definitive answers.
How to Get the Most From Your Results
Testing with more than one company is the single best way to build a clearer picture. Where two or three tests agree, you can be more confident. Where they diverge, you’ve identified the parts of your ancestry that current science simply can’t resolve with precision.
You can also upload your raw DNA data from one company to third-party tools like GEDmatch, which apply different reference panels and algorithms to the same data. This gives you additional perspectives without paying for a second test.
Pay more attention to the larger percentages in your results and less to the small ones. A result of 45% West African is meaningful. A result of 2% Scandinavian may be real, or it may be statistical noise from the algorithm struggling to classify an ambiguous DNA segment. And remember that ethnicity estimates update over time as companies expand their reference panels, so the results you see today may shift in a year as more people test and the underlying data improves.