MyHeritage DNA results are generally accurate at the level of raw genotyping, where the chip correctly reads common genetic variants well over 99% of the time. Where accuracy starts to slip is in the layers of interpretation built on top of that raw data: ethnicity estimates, health-related variant calls for rare mutations, and distant relative matching all carry meaningful margins of error. Understanding which parts of your results you can trust and which deserve skepticism depends on what kind of result you’re looking at.
How the DNA Chip Actually Works
MyHeritage, like most consumer DNA companies, uses an SNP genotyping array (often called a “chip”) to read your DNA. Rather than sequencing your entire genome, the chip checks hundreds of thousands of pre-selected spots across your DNA. Think of it as taking a standardized quiz at specific locations rather than reading the whole book. For the common genetic variants the chip is designed to detect, accuracy is excellent. A study evaluating SNP-chip data from the UK Biobank found that sensitivity, specificity, and precision were all above 99% for over 108,000 common variants.
The trouble starts with rare variants. That same study found that for variants occurring in fewer than 1 in 100,000 people, precision dropped dramatically, with only about 16% of those rare calls confirmed when checked against sequencing data. Nearly every individual tested had at least one rare pathogenic variant that was incorrectly genotyped by the chip.1bioRxiv. Assessing the analytical validity of SNP-chips for detecting very rare pathogenic variants: implications for direct-to-consumer genetic testing This matters because the variants most people worry about, the ones linked to serious diseases, tend to be rare. The chip performs beautifully on the easy stuff and struggles with the things that carry the highest stakes.
When Raw Data Gets Uploaded to Third-Party Health Tools
Many MyHeritage customers download their raw data file and upload it to third-party tools like Promethease, Codegen, or similar services that scan for health-related genetic variants. This is where a striking problem emerges. A clinical study that re-tested variants flagged in raw data from direct-to-consumer genetic tests found that 40% of reported variants were false positives: the chip said a mutation was present, but confirmatory clinical-grade testing showed it wasn’t there at all.2PubMed Central. False-positive results released by direct-to-consumer genetic tests highlight the importance of clinical confirmation testing for appropriate patient care
That 40% figure sounds alarming, and it should be. The explanation lies in the difference between what the chip was designed for and how the raw data gets used. SNP arrays are built to genotype common variants efficiently for ancestry and relative-matching purposes. They aren’t clinical diagnostic tools. When third-party services mine that raw data for rare pathogenic mutations the chip wasn’t optimized to detect, you get a high rate of errors. If you see a scary health result from uploading your MyHeritage raw data somewhere, do not panic and do not make medical decisions based on it. Clinical confirmation through your doctor using medical-grade sequencing is the only responsible next step.
How Ethnicity Estimates Are Built
Your MyHeritage ethnicity estimate isn’t just a readout of your DNA. It’s a statistical comparison between your genetic profile and a set of reference populations the company has assembled. The algorithm looks at patterns across your genotyped variants, compares them against these reference groups, and estimates what proportion of your ancestry likely traces to each region. This process involves a lot of modeling, and the results are best understood as educated estimates rather than hard measurements.
Several factors shape how accurate that estimate ends up being. The size and diversity of the reference panel matters enormously. If a company has sampled thousands of people with well-documented ancestry from, say, the British Isles, its ability to distinguish English from Scottish heritage will be stronger than its ability to parse fine-grained differences in West African or Southeast Asian ancestry, where reference samples may be thinner. Research has shown that imputation accuracy for common variants improves as reference panel size increases, but the benefit depends on how well the panel matches the population being studied.3Briefings in Bioinformatics. Genotype imputation and reference panel: a systematic evaluation on haplotype size and diversity
This is why your ethnicity percentages can shift when MyHeritage updates its algorithm or adds new reference populations. A change from “42% Eastern European” to “38% Eastern European and 6% Baltic” doesn’t mean your DNA changed. It means the company refined its comparison groups or improved its statistical model. These updates generally make results more granular, but they can be disorienting if you took the original numbers as fixed truths.
Why Different Companies Give Different Ethnicity Percentages
If you’ve tested with MyHeritage and also with AncestryDNA or 23andMe, you’ve probably noticed the ethnicity breakdowns don’t match perfectly. This is normal and expected. A comparison of genotyping platforms found that only about 16 to 18% of the SNPs with reported genotypes are shared across all three major companies. Each company selects its own set of genetic markers based on its own strategy, and the unique SNPs each platform chooses have different frequencies, ethnic selectivities, and chromosomal locations.4bioRxiv. SNP Selection and Concordance in Consumer Genetics Testing
On top of different chips, each company uses different reference populations and different algorithms. They also carve the world into different overlapping geographic regions, making an exact comparison impossible. That said, the same study found that where SNPs overlap across platforms, the genotype calls are highly concordant, with only two SNPs showing discordant calls out of over 110,000 shared markers. And when the overlapping regional definitions are accounted for, the companies produce broadly congruent results for the major contributors to your ancestry.4bioRxiv. SNP Selection and Concordance in Consumer Genetics Testing
So the raw genotyping agrees well across companies. The disagreements you see in ethnicity percentages are almost entirely a product of different reference panels, different regional definitions, and different statistical models. If MyHeritage says you’re 25% Scandinavian and 23andMe says you’re 18% Scandinavian and 9% broadly Northwestern European, those aren’t contradictory results. They’re two slightly different ways of carving up the same genetic signal.
Consistency Between Identical People
One useful way to gauge accuracy is to test identical twins, who share virtually all their DNA. A study that did exactly this found that when twin pairs were tested by the same company, the concordance of ancestry results was high, with mean percent agreement ranging from about 94.5% to 99.2%.5PubMed Central. Consistency of Direct to Consumer Genetic Testing Results Among Identical Twins The small discrepancies that remained reflect the inherent randomness in how algorithms assign ancestry percentages, since even the same DNA input can produce slightly different proportional breakdowns depending on how the statistical model handles borderline regions.
The takeaway is that within any single company’s platform, results are quite reproducible. The technology is reading your DNA consistently. The wobble in your ethnicity percentages isn’t a reading error at the chip level; it’s imprecision in the statistical modeling that sits on top of the genotyping.
Relative Matching and DNA Shared Segments
One of the most popular features on MyHeritage is DNA matching, where the platform identifies other users who share segments of DNA with you and estimates how closely you’re related. This relies on detecting segments of identity-by-descent, which are stretches of DNA two people inherited from a common ancestor. For close relatives like siblings, parents, or first cousins, these shared segments are long and easy to detect. MyHeritage is quite reliable for confirming close family relationships.
Where things get murkier is with distant matches. The further back in time the common ancestor lived, the shorter the shared DNA segments become. And short segments are far more prone to false positives. Research on large-scale pedigree data has shown that more than 67% of detected segments shorter than 4 centimorgans are false positives, meaning the algorithm flags them as shared ancestry when they’re actually just coincidental pattern matches. Only segments longer than about 5 centimorgans have a negligible false-positive rate.6Molecular Biology and Evolution. Reducing Pervasive False-Positive Identical-by-Descent Segments Detected by Large-Scale Pedigree Analysis
This means that when MyHeritage tells you someone is a predicted fourth or fifth cousin based on a few small shared segments, there’s a real chance that “match” is noise rather than a genuine family connection. By contrast, a match showing 100+ centimorgans of shared DNA is almost certainly a real relative. The practical rule of thumb: the larger the total shared DNA and the longer the individual segments, the more you can trust the relationship prediction.
The European Ancestry Advantage
A persistent issue across all consumer DNA companies, including MyHeritage, is that the reference databases used to build both ancestry estimates and health-related tools are heavily skewed toward people of European descent. Research has documented that polygenic risk scores and other genomic prediction tools are several times more accurate for individuals of European ancestry than for people of other backgrounds.7PubMed Central. Clinical use of current polygenic risk scores may exacerbate health disparities This isn’t because the underlying genetics are simpler in European populations. It’s because the vast majority of large-scale genetic studies have been conducted on European-descent cohorts, so the comparison data is richest there.
For ancestry estimation, this means MyHeritage can typically distinguish between, say, Irish and Italian heritage with reasonable precision, but may struggle to differentiate between, say, different West African or Indigenous American populations with the same level of detail. For health-related interpretations drawn from raw data, the disparity can be even more consequential. Individuals of non-European ancestry may not benefit equally from genomic tools because of this underrepresentation in the studies that inform those tools.8PubMed Central. Importance of Including Non-European Populations in Large Human Genetic Studies to Enhance Precision Medicine
MyHeritage has been expanding its reference populations over time, and its user base includes significant representation from certain non-European communities, particularly Jewish populations, where the company has historically had a strong user presence. But the structural bias in the field’s underlying research data remains a real limitation for anyone whose ancestry falls outside the well-studied European window.
Why Admixed Populations Get Fuzzier Results
If your ancestry is a mix of multiple continental or regional backgrounds, your ethnicity results will generally be less precise than those of someone whose ancestors all came from the same geographic area. This isn’t a flaw specific to MyHeritage; it’s a mathematical reality of how ancestry algorithms work. The reference populations used for comparison are typically built from individuals with relatively “clean” ancestral backgrounds from a single region. When your DNA is a blend, the algorithm has to decompose overlapping signals, and the uncertainty in each percentage estimate increases.
Research examining admixture-analysis tools in complex mixed populations has found that standard genetic tools and commonly used ancestral reference populations become less appropriate as societies grow more genetically complex.9PubMed Central. Putting RFMix and ADMIXTURE to the test in a complex admixed population If you’re a person with, say, African, European, and Indigenous American ancestry combined, the algorithm may confidently identify the broad continental components but misattribute the finer regional details. You might see percentages shift notably between updates as MyHeritage refines how it handles mixed-ancestry profiles.
Sample Quality and How Your Swab Affects Results
Something most people overlook is that the quality of the DNA sample you provide matters. MyHeritage uses cheek swabs to collect cells containing your DNA. If the swab is done poorly, if you ate or drank something shortly before, or if the sample degrades during shipping, the DNA may not perform as well on the genotyping array. Research comparing DNA from buccal swabs to blood-derived DNA found that degradation of swab-collected DNA affects both the total yield and the performance of genotyping, and that results can be seriously compromised if the sample quality isn’t adequate.10PubMed Central. Evaluation of quality of DNA extracted from buccal swabs for microarray based genotyping
In practice, MyHeritage and other companies have quality-control filters that catch most poor samples and ask you to resubmit. But borderline samples that pass quality control with marginal DNA quality could produce slightly noisier genotype calls, which would ripple into less reliable ancestry estimates and match results. Following the kit instructions carefully, avoiding food and drink for at least 30 minutes before swabbing, and mailing the sample promptly all improve your chances of a clean result.
What MyHeritage Does Well and Where to Be Skeptical
It helps to break down the different types of results by how much confidence you should place in each:
- Close family matches: Highly reliable. Parent-child, sibling, and first-cousin relationships are confirmed accurately by the amount of shared DNA.
- Broad continental ancestry: Generally reliable. If MyHeritage says you have significant European and West African ancestry, that’s almost certainly correct.
- Fine-grained regional breakdowns: Treat as estimates. The difference between “32% Greek” and “28% Italian” is within the margin of uncertainty for most people.
- Distant cousin matches: Increasingly unreliable past third cousins, especially when based on short shared segments under 5 centimorgans.
- Health variants from raw data uploads: Highly unreliable for rare mutations. Never act on a health finding from raw DTC data without clinical confirmation.
MyHeritage’s Database Size and What It Means for Matching
One factor that distinguishes DNA companies from each other is the size of their user database. A larger database means more potential matches, which is particularly useful for genealogical research. MyHeritage has built a sizable user base, especially strong in Europe and among people of Jewish descent. If your goal is connecting with living relatives for family tree research, the company where your relatives are most likely to have tested matters more than any technical advantage in the genotyping chip itself.
Database size also indirectly affects ethnicity estimates, because companies often use their own customer data (with permission) to build and refine reference populations. A company with more customers from a particular region can develop finer-grained regional categories for that area. This is one reason ethnicity results evolve over time and vary between companies. MyHeritage may offer more detailed breakdowns in regions where its user base is strongest, while AncestryDNA might have better resolution in areas where its customer base dominates, like North America.
For genetic genealogy specifically, many experienced researchers test with one company and then transfer their raw data to others, including MyHeritage, to cast the widest possible net for matches. MyHeritage accepts uploads from competing platforms, which has helped it grow its matching database. The genotyping accuracy of transferred data depends on how much overlap exists between the original chip and MyHeritage’s chip, which as noted earlier hovers around 16 to 18% of shared SNPs when comparing across all major platforms. Matches found through transferred data are still useful, but the shared-segment estimates may be slightly less precise than for kits processed natively on MyHeritage’s own chip.
Updates and Why Your Results Keep Changing
If you tested with MyHeritage a few years ago and check your results today, you may notice your ethnicity percentages have shifted. This happens because the company periodically updates its reference panels and analytical models. Each update incorporates more data, adds new reference populations, and sometimes restructures how regions are defined. A region that was once labeled “Balkan” might get split into “Greek” and “South Slavic” subgroups, or two previously separate categories might merge when the company determines it can’t reliably distinguish between them.
These updates generally improve accuracy in the aggregate, but they can feel disorienting on an individual level. Some customers feel strongly attached to a particular ethnicity percentage and are upset when it changes. It’s worth remembering that ethnicity estimates were never precise enough to warrant that level of attachment. The broad strokes of your results are stable; the small percentages are soft numbers that will keep shifting as the science and the data improve. If your results show 3% of something you can’t explain, it may be real admixture, or it may be statistical noise that disappears in the next update.