One person submitted samples to three DNA testing companies and received three different ancestry profiles. Was this fraud? An error? Neither — it was three slightly different answers to three slightly different questions.
A story that appears regularly in online discussions: a person takes ancestry tests from three different companies — Ancestry, 23andMe and MyHeritage — and receives three different results. One shows 40 percent Scandinavian ancestry, another 22 percent, a third 31 percent. Or one test shows 15 percent Ashkenazi Jewish ancestry, another shows 8 percent, a third shows 19 percent. 'This is all meaningless', the frustrated user concludes. Or: 'The test showed that my sister and I share only 23 percent of our DNA, but we are full siblings. Can that be right?'
In neither case is there fraud or error. There are several fundamental features of how DNA tests work that manufacturers do not always explain clearly enough — and which are critically important to understand for anyone using DNA testing for genealogy, finding relatives or planning to have a child.
Modern commercial DNA tests use a method called SNP genotyping. SNP stands for Single Nucleotide Polymorphism. To understand what this means: your genome — the full set of your DNA — contains roughly three billion pairs of chemical letters. At most positions in the genome, every human has the same letter. But at specific positions, different people carry different letters: one person has an A where another has a G, for instance. These variable positions are the SNPs. They are, in effect, the genetic markers that make individuals different from one another.
Rather than reading your entire genome (which would be expensive), a commercial test examines several hundred thousand to a few million of these variable positions — a selection chosen in advance by the company. Different companies choose slightly different sets of positions. This is the first source of variation between results.
The resulting data — your SNP profile — is then compared against reference databases. A reference database is a collection of DNA samples from people the company has classified as representatives of particular groups — 'British', 'Scandinavian', 'Ashkenazi Jewish' and so on. An algorithm then calculates how similar your profile is to each of those groups. Different companies use different reference databases, with different samples and different sizes. This is the second source of variation.
Imagine company A has built its 'Scandinavian' reference group primarily from people in Norway and Sweden with several generations of documented ancestry from those countries. Company B has included people from Denmark, Finland and parts of the Baltic region as well. That is already a different database — and it will produce slightly different results for the same genome. The two companies are not wrong. They are answering slightly different questions.
The size of the database also matters. If a company's reference group for 'Polish' ancestry contains only three hundred people, the algorithm will work less accurately than if it contains thirty thousand. Small databases do not capture the real genetic diversity within a population. This problem is particularly noticeable for smaller, historically isolated or commercially under-represented populations: Mizrahi Jews, Kurds, Berbers, many South Asian communities. Their results tend to be less precise than for Western European populations, which are well represented in commercial databases.
Many users are puzzled when a company updates its algorithm and their ancestry results change noticeably — sometimes by ten or fifteen percentage points in a category. This does not mean the company made a mistake earlier or is making one now. It reflects a genuine scientific challenge.
To understand this, it helps to know a concept called linkage disequilibrium. This refers to the fact that genetic variants located physically close together on the same chromosome tend to be inherited together, generation after generation, because the biological process that shuffles chromosomes between generations rarely separates them. Over time, populations accumulate distinctive combinations of variants — patterns that are characteristic of their particular history. These patterns differ between populations, and identifying them is part of how ancestry algorithms work.
The complication is that modern human populations are themselves the products of historical mixing, and the boundaries between genetic clusters are gradual rather than sharp. When a company refines its algorithm or expands its reference database, the boundaries it draws between clusters shift — and so do individual results. This is not failure; it is science improving. Current academic research is moving towards multi-ancestry models that try to account for the fact that a person's genome is a mosaic of components from different historical sources, rather than assigning each segment a single neat ethnic label. These models are more accurate — but harder to summarise in a simple percentage.
This is one of the most frequently asked questions about DNA tests: 'My sister and I are full siblings, but the test says we share only 23 percent of our DNA. Is that normal?' The answer requires a brief explanation of how inheritance works.
Both you and your sister received half of your DNA from your mother and half from your father. But you did not receive the same half. When a parent produces a reproductive cell — an egg or a sperm — each of the parent's chromosome pairs is first split and reshuffled in a process called recombination. Think of it as cutting two decks of cards and mixing them together in a new order. Which segments came from which grandparent is determined randomly, independently for each reproductive cell. The result is that you and your sister received different random selections from the same parental genomes.
On average, full siblings share about 50 percent of their DNA. But there is a natural range around this average. The actual range for full siblings is roughly 38 to 61 percent of shared DNA. A result of 23 percent would fall below the normal range for full siblings, which would suggest either a half-sibling relationship (one shared parent rather than two) or, less likely, a technical issue with the test. Either way, it is a reason to investigate further — not a reason to assume there is a problem with the test itself.
'Genetic compatibility' between partners: what this actually means
In the context of choosing a donor or co-parent, the question of 'genetic compatibility' sometimes arises. It is important to separate two fundamentally different meanings of this phrase.
The first meaning is ethnic or ancestral similarity: how similar two people are in their ancestral background. This is information about geographic origin, but it does not predict a child's health or appearance, and it is not a measure of biological suitability for parenthood. The second — and medically relevant — meaning is carrier compatibility: whether both prospective parents carry a harmful recessive variant in the same gene. If both parents carry such a variant, each child they have together has a 25 percent probability of inheriting two copies and developing the associated condition.
A carrier screening test — which examines specific, medically documented disease-causing variants in genes associated with conditions such as cystic fibrosis, spinal muscular atrophy, Tay-Sachs disease and several hundred others — provides genuinely useful information for reproductive planning. It is incomparably more relevant to a child's health than any ethnic ancestry profile. These are two entirely different tests answering entirely different questions.
A few practical conclusions from everything above. First: ancestry percentage breakdowns are estimates, based on the specific reference database of a specific company. They are useful as a starting point for genealogical research, but they are not exact measurements of your genetic 'composition'. Second: different companies giving different results is normal — it does not mean that one of them is wrong. Third: for medical purposes — planning reproduction, assessing disease risk — a specialised clinical genetic test is needed, not a consumer ancestry product. Fourth: the percentage of shared DNA with a relative is a range, not a fixed number, and the range is wider than most people expect.
DNA tests are a genuinely valuable tool for genealogy and for understanding one's origins. They work, and they provide real information. But that information has a specific nature and specific limitations. Different results from different companies are not fraud and not error. They reflect the fact that different companies use different algorithms and different reference databases to answer similar, but not technically identical, questions. Understanding these nuances does not reduce the value of the tests — it allows them to be used more thoughtfully and more accurately.
SNP (Single Nucleotide Polymorphism) — a position in the genome where different people carry different genetic letters. The basic unit of analysis in commercial DNA tests.
Linkage disequilibrium — the tendency for genetic variants located close together on the same chromosome to be inherited together rather than independently. Different populations have different patterns of linkage disequilibrium, which helps ancestry algorithms identify ancestral origins.
Reference database — a collection of DNA samples from people classified as representatives of particular populations. The quality and composition of the reference database directly determines the accuracy of a test's results.
Autosomal DNA — the DNA of the 22 pairs of non-sex chromosomes (chromosomes 1 to 22). This is what commercial ancestry tests primarily analyse. It is distinct from Y-chromosome DNA (inherited only through the paternal line) and mitochondrial DNA (inherited only through the maternal line).
Carrier screening — a clinical genetic test that determines whether a person carries a recessive disease-causing variant in a specific gene, without being affected by the condition themselves.
Thousands of people are already building families on their own terms.
Browse profiles