How DNA Ancestry Testing Works and What the Results Actually Mean
Published:
In brief
DNA ancestry tests compare hundreds of thousands of genetic markers from a saliva sample against reference databases of people with documented family origins, producing statistical estimates rather than definitive proof of heritage. Results differ between companies because reference panels vary in size and composition, and the 2025 bankruptcy sale of 23andMe showed that sending a sample also raises lasting data-privacy questions.
A direct-to-consumer DNA ancestry kit does not read a person's full genome. It scans a saliva sample for several hundred thousand specific positions in the DNA called single nucleotide polymorphisms, or SNPs — spots where human genetic code commonly varies from person to person, according to the U.S. National Library of Medicine's MedlinePlus Genetics resource.
Three Kinds of Tests, Three Kinds of Ancestry
MedlinePlus Genetics describes three main testing approaches. Y-chromosome testing traces paternal-line ancestry and only works on male customers, because only fathers pass a Y chromosome to sons. Mitochondrial DNA testing traces the maternal line in both sexes, since mitochondria are passed down only from mothers. Autosomal DNA testing, the method used by most consumer kits such as AncestryDNA and 23andMe, examines the 22 non-sex chromosome pairs inherited equally from both parents and gives what MedlinePlus calls a more comprehensive picture of ancestry.
How a Percentage Breakdown Gets Calculated
After the lab records a person's SNP pattern, a computer program compares it statistically to a 'reference panel' — a database of people with long, documented ancestry in a specific region. A peer-reviewed method described in a study published via the National Center for Biotechnology Information (iAdmix) uses a maximum-likelihood calculation: it estimates what mix of reference populations' allele frequencies would best explain the customer's own genetic pattern, treating each region as a proportion that should add up to the whole genome.
The same study found this approach performs reliably across continents but loses precision between closely related populations. When two reference groups have very similar allele frequencies — for example, two neighboring European populations — the software can split a person's European ancestry between them unpredictably, the researchers reported, because the genetic signals are nearly identical.
Why Two Companies Can Give Different Numbers for the Same Person
Reference-panel size and composition differ sharply between companies, which is one reason the same customer can get different regional breakdowns from different kits. According to genetic-genealogy researcher Leah Larkin, writing as The DNA Geek, AncestryDNA's reference panel averages 818 individuals per defined region, compared with roughly 320 for 23andMe, about 119 for MyHeritage and about 90 for FamilyTreeDNA. She also reported that AncestryDNA added 12,134 people to its panel in a 2022 update, making it nearly five times larger overall than 23andMe's, at the time the next-largest.
Average reference-panel size per ancestry region, by company
Company
Avg. individuals per region (approx.)
AncestryDNA
818
23andMe
320
MyHeritage
119
FamilyTreeDNA
90
MedlinePlus Genetics lists further sources of inconsistency: some world populations remain underrepresented in every company's database, historical migration means a genetic 'origin' is not necessarily where recent ancestors actually lived, and each test analyzes only a portion of a person's total genetic information.
What the Results Can and Cannot Tell You
Results are probability-based estimates, not a legal or definitive record of ethnicity or nationality.
A region listed in a report reflects genetic similarity to modern reference populations, not necessarily where ancestors lived historically.
Different companies can legitimately produce different percentages for the same person because their reference databases differ.
Tests do not capture cultural identity, upbringing or social relationships — MedlinePlus states ancestry results are 'just one part of a person's story.'
What Happens to the Sample After You Send It
Consumer genetic data is treated as unusually sensitive because, as the U.S. Federal Trade Commission put it in a 2023 consumer alert, DNA data is sensitive 'because it's about who you are.' The same FTC alert cited an enforcement case against the company Vitagene, which left detailed health reports of nearly 2,400 consumers and raw genetic data of at least 227 others exposed online despite promising industry-standard security.
The FTC alert also notes a legal gap: the federal Genetic Information Nondiscrimination Act bars genetic discrimination in employment and health-insurance decisions, but that protection does not extend to disability insurance or long-term-care insurance.
Ownership of a DNA database can also change. 23andMe filed for Chapter 11 bankruptcy in March 2025 with the genetic data of roughly 15 million customers at stake, according to HIPAA Journal. A bankruptcy court approved a $305 million purchase by TTAM Research Institute, a nonprofit led by 23andMe co-founder Anne Wojcicki, after an earlier bid from drugmaker Regeneron was topped. TTAM committed not to sell or transfer the genetic data again unless a future buyer adopts its privacy policies, pledged to set up a consumer privacy board within 90 days, and offered two years of free identity-theft monitoring to affected customers, per HIPAA Journal's reporting.
individuals should be empowered to have choice and transparency with respect to their genetic data.
— Anne Wojcicki, 23andMe co-founder, quoted by HIPAA Journal
Before buying a kit, the FTC advises consumers to read a company's privacy terms, confirm who can access or receive the data, and understand how it will be stored and protected — since, unlike a password, a person's DNA sequence cannot be changed once it has been shared.
Questions
How accurate are DNA ancestry test results?
Results are statistical estimates, not definitive proof. Accuracy depends on how large and diverse a company's reference database is, and a peer-reviewed method described via NCBI found the approach works well across continents but loses precision between closely related neighboring populations.
Why do AncestryDNA and 23andMe give different ethnicity percentages for the same person?
The two companies use reference panels of different size and composition. Genetic-genealogy researcher The DNA Geek reports AncestryDNA's panel averages 818 people per region versus about 320 for 23andMe, which can shift how the same DNA is categorized.
What's the difference between autosomal, Y-DNA and mitochondrial DNA tests?
Per MedlinePlus Genetics, Y-chromosome tests trace only the paternal line in males, mitochondrial DNA tests trace only the maternal line, and autosomal tests examine chromosomes from both parents, giving the broadest ancestry picture.
What happens to my DNA sample and data after I submit a test?
Companies may use de-identified data for research and must disclose this in their privacy terms, per the FTC. Ownership of a genetic database can also change hands, as shown when 23andMe's 15 million customer records were sold for $305 million during its 2025 bankruptcy, according to HIPAA Journal.
Is my genetic ancestry data legally protected from misuse?
The FTC notes that the Genetic Information Nondiscrimination Act bars genetic discrimination in employment and health insurance, but it does not cover disability insurance or long-term-care insurance.
This piece was prepared with artificial intelligence and passed an automated source check before publication. If you see an error, ask for a correction. Request a correction or reply · How we make this
Mitosis is the cell division that makes two identical copies of a cell for growth and repair, while meiosis is the specialized division that produces egg and sperm cells with half the usual chromosome number. Both processes copy a cell's DNA once, but mitosis divides once and meiosis divides twice, according to the U.S. National Institutes of Health.
On 15 October 2026, the General Conference on Weights and Measures (CGPM) voted to stop inserting leap seconds into Coordinated Universal Time (UTC), effective 20 May 2027. The change means UTC will run continuously and be allowed to drift up to one hour from Earth's actual rotation before any future correction, instead of the one-second adjustments used since 1972.
The General Conference on Weights and Measures voted on 15 October 2026 to stop adding leap seconds to Coordinated Universal Time (UTC), making the time standard continuous from 20 May 2027. The decision, adopted as Resolution 5 at the BIPM's 28th meeting in Versailles, France, replaces the one-second correction used since 1972 with a tolerance of up to one hour.
Japan's Meteorological Agency (JMA) issued and then withdrew, within about eight to twelve minutes, an emergency warning for a magnitude-9.9 earthquake on October 16, 2026. The actual quake, centred off Russia's Kamchatka Peninsula, was later confirmed by seismologists at magnitude 6.9, with JMA blaming a data-processing error rather than a cyberattack.
