How Accurate Is 23andMe?

23andMe is accurate at reading your DNA and at finding your genetic relatives. The lab reads hundreds of thousands of DNA markers at better than 99% accuracy, and its DNA Relatives matches are real, measured shared DNA. What is less exact, by design, is the ancestry composition breakdown and the health reports, which are estimates and screens rather than final answers.

In short, 23andMe has two solid layers and two "read with care" layers. The raw genotyping and the relative matching are dependable. The ethnicity percentages are educated estimates, and the health results are screening indicators, not diagnoses. This guide walks through each so you know how much weight to give it. If a listed relative leaves you unsure how you actually connect, the free relationship calculator names the link fast.

The Reading of Your DNA: Very Accurate

Like other consumer kits, 23andMe uses a genotyping microarray. It reads a fixed set of roughly 600,000 to 700,000 SNPs (spots where people commonly differ) rather than sequencing your whole genome. This is a well-established lab process, and per-marker accuracy is typically quoted at better than 99%. Samples that read poorly get flagged or re-run, which is why a small share of testers are asked to spit again.

The honest limit is the same for every genotyping test: it reads only the spots the chip targets. It is not whole-genome sequencing, so a rare private variant outside those spots simply is not read. That matters for edge-case medical questions but not for ancestry or relative-finding, which is what most people are after.

DNA Relatives: Reliable Measurement, Estimated Labels

23andMe's DNA Relatives compares your DNA to other testers and lists those who share segments with you. The shared DNA it reports is a real measurement in centimorgans (cM). If it says you and someone share a large block of DNA, you genuinely do. That measurement is trustworthy.

The estimate lives in the relationship label, because a single cM amount can fit more than one relationship.

Because ranges overlap, a mid-range match could carry several labels. Close relatives are the exception: a parent, full sibling, or first cousin sits in a range nothing else imitates, so those are essentially certain. For distant matches, use the label as a starting hypothesis and confirm it against your tree. The cousin calculator helps you test which relationship the shared DNA actually supports.

One structural note: 23andMe's DNA Relatives has historically capped how many matches it shows and requires opt-in, so your visible match list may be shorter than on a records-focused service. That is a product setting, not an accuracy problem.

Ancestry Composition: A Good Estimate, Not a Measurement

23andMe's ethnicity feature, Ancestry Composition, compares your DNA to reference populations and estimates how much of your DNA best matches each. Like every ethnicity estimate, it is exactly that, an estimate, and it behaves like one:

• Continental signals are strong. Broadly European, African, East Asian, and similar high-level results are usually reliable.
• Country and regional splits are softer. Neighboring populations share DNA, so a Germany-versus-France or Ireland-versus-Scotland split is blurry.
• Trace percentages are the least certain. A small single-digit region may be a real distant ancestor or statistical noise.
• Results update over time. When 23andMe expands reference data or refines its model, everyone gets re-estimated. A shifting number is the estimate sharpening, not your DNA changing.

23andMe lets you slide a confidence setting from speculative to conservative. Turning up confidence collapses uncertain small regions into broader, safer categories, a useful way to see which parts of your result the system is actually sure about. Our guide on DNA ethnicity accuracy explains why these numbers move.

Haplogroups: Accurate, But Narrow

23andMe also reports a maternal haplogroup (from mitochondrial DNA) and, for those who inherit a Y chromosome, a paternal haplogroup (from Y-DNA). These place you on deep, ancient branches of the human family tree and are reliably assigned.

The catch is scope. Your maternal haplogroup traces only your mother's mother's mother's line, and the paternal haplogroup traces only your father's father's father's line. They ignore the vast majority of your ancestors. They are accurate for what they describe, but they describe two threads out of thousands. Our explainer on mtDNA, Y-DNA, and autosomal testing shows how these test types differ.

Health Reports: Screens, Not Diagnoses

23andMe is notable for FDA-authorized health reports, but their accuracy has to be understood correctly, because this is where misreadings cause real harm.

The reports check for specific, named variants. For example, a BRCA report checks a limited set of variants most common in certain populations, not every possible BRCA variant. So a "variant not detected" result does not mean you have no risk; it means the specific variants tested were not found. Conversely, a flagged variant is a reason to talk to a doctor and get clinical confirmation, not a diagnosis on its own.

Read the health reports as a screening tool that can start a useful conversation with a clinician. Any result that would change a medical decision should be confirmed by a diagnostic test ordered through a healthcare provider. This is not a knock on 23andMe; it is how consumer genetic screens are meant to work.

Why Some Results Look "Wrong" (But Are Not)

Most complaints that "23andMe got it wrong" trace back to a handful of predictable misreadings, not a lab error. Recognizing them saves a lot of worry.

• A missing ancestor in the ethnicity map. If your great-grandparent was, say, Italian, but no Italian percentage appears, that is often normal. DNA passes down randomly, and after a few generations a single ancestor may leave too little detectable signal to label. The ancestor was real; the DNA simply thinned.
• A surprising close match. Occasionally a match reveals a family connection nobody discussed, such as a half-sibling or a different biological parent than expected. This is not the test malfunctioning. Close-relationship shared-DNA ranges are unmistakable, so a surprising close match is far more likely to be a true, previously unknown relationship than an error.
• Percentages that do not sum to your family story. You will not get a clean 25% per grandparent, because inheritance is random, not evenly divided. Uneven numbers are expected, not a defect.
• A number that changed. As covered above, updates re-estimate everyone. A shifted percentage is the model improving.

When a match itself is the surprise, the most useful next step is to work out exactly how the shared DNA could connect you, and the cousin calculator helps you test the possibilities against your known tree.

How 23andMe Compares on Accuracy

Raw-reading accuracy is broadly similar across the major genotyping kits because the chip technology is comparable. The differences are elsewhere: 23andMe is stronger on haplogroups and health features, while a records-focused rival may surface more distant cousin matches thanks to a larger database. As of recent reporting, 23andMe's database is sizable but smaller than the largest genealogy-focused services, which mainly affects how many distant matches appear, not the accuracy of the ones you get. For direct comparisons, see AncestryDNA vs 23andMe and MyHeritage vs 23andMe.

Ancestry Composition vs the Match List: Two Different Accuracies

A useful mental split when you open your 23andMe results is that the ethnicity map and the match list are accurate in completely different ways, and confusing them is the root of most disappointment.

The match list is accurate in the hard sense: it measures real shared DNA, and if two people overlap on a long segment, they truly descend from a common ancestor. There is no interpretation involved in whether the DNA overlaps. That is why, for genealogy, the match list is the tool you act on.

The Ancestry Composition map is accurate in a softer, statistical sense: it is a good estimate of how much your DNA resembles various reference populations. It is genuinely informative at the continental level and genuinely uncertain at the country level. It is not "wrong" when it is fuzzy; fuzziness is built into estimating overlapping populations.

So the honest summary is: trust the match list to find and confirm relatives, and read the ethnicity map as an informed pointer to where your family broadly came from. Treating the soft map as if it were as hard as the match list is the single most common way people end up feeling misled by an accurate test.

Getting an Accurate Result

• Collect the sample properly. No food, drink, or smoking for 30 minutes before, and reach the fill line. A weak spit sample is the usual cause of a genuine failure.
• Trust matches over the map. For finding relatives, DNA Relatives is the accurate, actionable feature.
• Use the confidence slider. Set Ancestry Composition to a higher confidence to see which regions the model is truly sure about.
• Confirm health flags clinically. Treat any meaningful health result as a prompt for a doctor, never as a final answer.
• Download your raw data. Your raw DNA file is yours to keep. Downloading it lets you upload to other databases later and widen your match pool without buying another kit, a good habit even if you have no immediate plan for it.

Autosomal vs Deep-Line: What 23andMe Can and Cannot Tell You

It helps to know which questions 23andMe is built to answer. The bulk of its power comes from an autosomal test, DNA inherited from all your lines, which is excellent for finding cousins out to several generations and for a broad ancestry picture. What autosomal testing does poorly is trace one specific line far back, because the signal from any single ancestor fades quickly with each generation.

The haplogroups fill part of that gap for two lines only, your direct maternal line and, if applicable, your direct paternal line. But they will not tell you about, for instance, your mother's father's mother's line, which is neither of those two threads. So if your real goal is to prove a specific deep line, such as a surname study running back many generations, 23andMe is not the ideal tool; a dedicated Y-DNA or mtDNA test from a specialist service is. For everyday questions, who are my cousins, where did my family broadly come from, 23andMe is well suited.

FAQ

Is 23andMe accurate for ancestry?

At the continental level, yes, it is quite accurate. Specific country and regional percentages are educated estimates that can shift with updates, and small trace regions carry the most uncertainty. The relative-matching feature is the most reliable part for genealogy.

How accurate are 23andMe health reports?

They accurately test for specific named variants, but they are screens, not diagnoses. A "not detected" result does not rule a condition out, because only certain variants are checked. Confirm any meaningful result with a clinical test through a doctor.

Is 23andMe accurate for finding relatives?

Yes. DNA Relatives measures real shared DNA, and close relatives are matched with near certainty. Distant matches are also real, though the exact cousin label is a best estimate from overlapping ranges.

Why did my 23andMe ancestry results change?

Because 23andMe periodically re-estimates everyone as it adds reference data and improves its model. Your DNA did not change; the estimate became more refined. Shifts of a few percentage points are normal.

How accurate are 23andMe haplogroups?

Very accurate for what they measure, which is your direct maternal line and, if applicable, your direct paternal line. They reliably place those two lines on deep ancestral branches but say nothing about the rest of your ancestry.

The Bottom Line

23andMe reads your DNA accurately and finds your genetic relatives reliably, with close family matched at near-certainty. Its Ancestry Composition is a solid continental estimate that gets softer for countries and trace regions, and its health reports are screening tools to confirm clinically rather than diagnoses. Understand which layer you are looking at and the results are genuinely useful. When a listed relative leaves you wondering exactly how you connect, the free family relationship calculator turns shared DNA and a bit of family knowledge into the right relationship name.