When you spit into a tube and mail it off to AncestryDNA or 23andMe, what actually happens to your DNA? The answer reveals why ancestry testing is simultaneously one of the most accessible and one of the most misunderstood areas of personal genomics. Your ethnicity pie chart is not a fixed truth — it is a statistical estimate built on reference populations, algorithms, and probabilities that shift every time a company updates its model.
How Autosomal DNA Testing Works
Consumer ancestry tests analyze autosomal DNA — the 22 pairs of chromosomes you inherit from both parents. Each test uses a microarray chip that reads between 600,000 and 700,000 single nucleotide polymorphisms (SNPs) scattered across your genome. These SNPs are compared against reference panels: collections of DNA samples from people with known ancestry going back multiple generations in a single geographic region.
Your DNA is broken into short segments called "windows" — typically 100–300 SNPs each — and each window is compared to the reference panel. A statistical model assigns a probability that each window came from a particular population. The final ethnicity estimate is the aggregate of those probabilities across your entire genome. This is why results are expressed as percentages with confidence intervals, not certainties.
AncestryDNA maintains the largest consumer database with over 25 million users, which gives it the largest reference panel and — in theory — the most statistically robust estimates. 23andMe, with over 14 million genotyped customers, covers more than 2,000 geographic regions, the most granular breakdown among major consumer tests. MyHeritage covers 2,114 regions and excels at European and Jewish ancestry breakdowns. FamilyTreeDNA, with a smaller database but deeper genealogical tools, is favored by serious family historians.
Why Different Companies Give Different Results
The single most common question from ancestry testers: "Why does my 23andMe result say 22% Italian while AncestryDNA says 14%?" The answer lies not in errors but in methodology. Each company selects its own reference populations, defines its own geographic boundaries, and uses proprietary algorithms to assign ancestry probabilities.
Reference populations are the foundation of every estimate. If a company's reference panel for "Ireland" includes 500 samples from Dublin and Cork but none from Galway, someone with deep Galway ancestry may be misassigned to "Scotland" or "England" simply because the reference panel doesn't represent their actual population. This is particularly problematic for underrepresented populations. Indigenous American, African, and Southeast Asian reference panels tend to be significantly smaller than European ones, leading to less precise estimates for those ancestries.
Algorithm updates also play a major role. AncestryDNA typically updates its ethnicity estimates annually, and users frequently see percentages shift by 5–15% — sometimes entire regions appear or disappear. This does not mean the earlier estimate was "wrong." It means the reference panel grew and the statistical model improved. The American Journal of Human Genetics documented in 2020 that reference panel composition is the single largest factor affecting ethnicity estimate accuracy across platforms.
DNA Matching and Building Your Family Tree
While ethnicity estimates grab headlines, DNA matching is where ancestry testing delivers its most concrete value. When you test, the company compares your DNA segments against every other user in its database and identifies people who share significant stretches of identical DNA. The length of shared DNA, measured in centimorgans (cM), predicts how closely you are related.
A parent-child pair shares approximately 3,400 cM across their genome. Full siblings share around 2,600 cM. First cousins share roughly 850 cM. Second cousins share about 230 cM. Beyond third cousins (under 50 cM), the shared DNA becomes small enough that relationships become harder to predict with confidence. The International Society of Genetic Genealogy maintains widely referenced cM range charts that genealogists use to interpret match results.
The power of DNA matching lies in triangulation — when three or more people all share the same DNA segment and all trace their ancestry to the same common ancestor, you have genetic proof of that ancestral connection. Serious genealogists build chromosome maps, marking which segments came from which ancestors, creating a DNA-based family tree that supplements (and sometimes corrects) paper-trail genealogy.
Y-DNA and mtDNA for Deep Ancestry
Autosomal tests only reach back about 5–7 generations before the shared DNA becomes statistically indistinguishable from random. For deeper ancestry, Y-DNA and mitochondrial DNA (mtDNA) tests provide tools that reach back thousands or tens of thousands of years.
Y-DNA testing analyzes the Y chromosome, which is passed almost unchanged from father to son. Because surnames in many cultures follow the paternal line, Y-DNA testing can confirm whether two men with the same surname share a common paternal ancestor. FamilyTreeDNA offers Y-DNA tests at 37, 111, and 700 STR marker resolutions, with the Big Y-700 test providing the most comprehensive analysis of the Y chromosome.
mtDNA testing traces the direct maternal line through mitochondrial DNA, which is inherited exclusively from the mother. While mtDNA mutates slowly, making it less useful for recent genealogy, it maps deep ancestral migration paths, allowing individuals to identify their maternal haplogroup and trace it back to ancient origins. The Genographic Project, a multi-year National Geographic initiative, used mtDNA extensively to map human migration out of Africa and across the globe.
Third-Party Tools: GEDmatch and Beyond
Each DNA testing company maintains a walled garden. 23andMe users can only match with other 23andMe users. AncestryDNA matches are limited to its own database. GEDmatch solves this problem by accepting raw DNA data uploads from all major testing companies, creating a cross-platform database where you can find relatives regardless of which company they tested with.
GEDmatch also provides advanced ethnicity calculators — Eurogenes, Dodecad, HarappaWorld, MDLP — developed by independent researchers. These calculators often produce more granular breakdowns than commercial tests, particularly for mixed ancestry and underrepresented populations. Other third-party tools like DNA Painter allow you to map which DNA segments came from which ancestors, building a visual chromosome map of your genetic inheritance.
Testing Strategy: Start With the Oldest Generation
Genetic genealogists universally recommend testing the oldest living relatives first. Your parents each carry approximately twice as much DNA from your shared ancestors as you do. Your grandparents carry roughly four times as much, and so on. Every generation that passes without being tested represents genetic information permanently lost.
Test your oldest relatives with autosomal DNA tests, ideally on AncestryDNA (largest matching database) or 23andMe, then upload those raw data files to MyHeritage, FamilyTreeDNA, and GEDmatch for cross-platform matching. For deep paternal line questions, pair autosomal results with a Y-DNA test from FamilyTreeDNA. For maternal line questions, an mtDNA test captures the matrilineal heritage that autosomal testing might miss.
Ancestry testing is a living science — your results are a snapshot of what reference populations and algorithms can tell you today, not a definitive biological truth. As databases grow and algorithms improve, the picture sharpens. The DNA itself doesn't change, but our ability to read it improves every year.