What your 23andMe raw data reveals about your health

A plain-language explainer · Updated June 2026

23andMe & raw DNA

What raw data actually is — and what it can't do

Before the categories, it helps to understand what you are actually reading. When you download your raw data — whether from 23andMe or AncestryDNA — you get a plain text file: a long table of marker IDs and the two letters (your genotype) found at each one. That's it. No interpretation, no verdicts, just positions and letters.

The critical thing to grasp is how little of your genome that table covers. Your DNA is roughly three billion base pairs long. A consumer genotyping chip reads only a few hundred thousand pre-selected positions — a tiny sampled fraction, chosen because they're informative for ancestry or common traits, not because they form a complete health picture. Most of your genome is never read at all. So when a variant doesn't appear in your file, that very often just means the chip never looked at that spot — not that you don't carry it.

That single fact reshapes everything below. A variant in your raw data is an association drawn from population research — a statistical link between a marker and an outcome — not a verdict about you personally. Most health-relevant traits are polygenic: shaped by hundreds or thousands of variants acting together, plus diet, sleep, environment, and chance. And because these chips are built for scale rather than diagnosis, each individual reading carries a small but real error rate. That's why confirmation matters, and why we repeat the same caveat throughout: this is research-grade, curiosity-grade information, not a clinical test.

With that framing set, here are the categories of what people actually look for.

Category 1 — Carrier status and serious risk variants

This is the category people handle with the most anxiety, and the one that most needs care. It covers two overlapping things: carrier status for recessive conditions, and markers linked to serious health risks like hereditary cancer or clotting disorders.

Carrier status is about recessive conditions — conditions that typically only appear when someone inherits two copies of a variant, one from each parent. You can carry a single copy, be perfectly healthy, and never know unless you look. This matters most for family planning, since two carriers can have an affected child. A common example readable from raw data is hemochromatosis, linked to variants in the HFE gene, which affects how the body absorbs iron. We cover what those markers mean in checking hemochromatosis (HFE) in your 23andMe raw data.

Then there are the serious risk variants — and here the incomplete-coverage problem becomes safety-critical. BRCA is the clearest case. 23andMe tests only a small, specific set of BRCA1/BRCA2 variants — chiefly a few founder mutations common in people of Ashkenazi Jewish descent. There are thousands of other cancer-risk variants in those genes that the chip never checks. So a "variant not detected" result is not reassurance and not a clean bill of health — it simply means the handful of positions tested happened to be negative. We unpack this fully in what 23andMe raw data does and doesn't tell you about BRCA.

The same logic applies to clotting risk. Variants like Factor V Leiden can appear in raw data, but the absence of a clotting variant in your file does not rule out an inherited clotting disorder.

A related but distinct case is complex, polygenic risk — heart disease, type 2 diabetes, and late-onset Alzheimer's. Here no single marker decides anything; the outcome is shaped by many variants plus lifestyle. The best-known example is APOE4, associated with higher Alzheimer's risk. Carrying it raises a statistical probability — it does not mean you will develop the disease, and many people with the variant never do, while plenty without it still do. Because a dementia-risk marker is emotionally heavy and easy to misread, this one especially deserves the counsellor-and-context treatment rather than a solo reading of your file. We cover it carefully in what APOE4 in your 23andMe raw data means. The mental model across this whole category stays the same: risk, not destiny.

Category 2 — Pharmacogenomics: how you metabolise drugs

Some of the most practical-sounding information in your file is pharmacogenomic — how your genetics may influence the way you process certain medications. Genes in families like CYP2D6 and CYP2C19 encode enzymes that metabolise common drugs, and variants can make someone a faster or slower metaboliser of a given medicine. This can relate to how some painkillers, antidepressants, or blood-thinners are handled by the body — see what CYP2C19 in your 23andMe raw data means for a worked example.

It is genuinely useful context — but this is the category where the "never act on raw data" rule is most important, because the stakes are a real prescription. Do not start, stop, or change any medication based on a raw-data file. The results can be wrong, they cover only a fraction of the variants that matter, and dosing is a clinical decision. If a result seems relevant, bring it to your prescriber, who can order validated clinical pharmacogenetic testing before it informs anything.

Category 3 — Everyday traits and wellness

The lowest-stakes and often most fun category covers everyday traits and wellness — the small ways genetics shows up in daily life. These rarely carry the weight of the serious markers, but they can be a useful nudge toward paying attention to something, and the associations here are generally better understood.

• Caffeine metabolism — variants linked to CYP1A2 are associated with how quickly you clear caffeine, which some people find lines up with how coffee affects their sleep. See CYP1A2 and caffeine in your raw data.
• Alcohol flush — the ALDH2 variant behind the alcohol-flush reaction is one of the clearest gene-to-trait links there is, and it carries real health weight beyond the visible flush.
• Lactose tolerance — a marker near the LCT gene is associated with whether you keep digesting lactose comfortably into adulthood; more in is lactose intolerance genetic?.
• Folate metabolism — variants in the MTHFR gene are among the most-searched in all of consumer genetics, often more myth than meaning. We separate signal from noise in what MTHFR in your 23andMe raw data actually means.

Even here, the same caveats apply: these are population-level associations, not rules about you. They are gentle prompts to notice your own patterns — and the kind of insight that is far more interesting when it sits next to your actual sleep, diet, and lab data rather than floating alone.

Category 4 — Ancestry-linked health context

Because DNA is inherited along family and population lines, some variants are more common in particular ancestral groups. This is the fourth thing people look for, and it's easy to misunderstand. The BRCA founder mutations 23andMe tests are the classic example — they were chosen partly because they occur more often in one ancestral population, which is exactly why a negative result tells you so little if you're outside it.

Handled well, ancestry-linked context is genuinely useful: knowing that a certain hereditary condition runs more commonly in your background can help you and a clinician decide which clinical screening is worth pursuing. Handled badly, it becomes a false sense of security ("that condition isn't common in my group, so I'm fine") or unfounded alarm. Treat it as one input a genetic counsellor can weigh — background context, never a diagnosis — and remember the coverage limits apply here just as much as anywhere.

How to actually look these up

Knowing the categories is one thing; reading your own file is another. Your raw data is just a table of marker IDs and genotypes — unreadable until a tool translates it into plain language. The quickest privacy-friendly way to start is our free DNA explorer, which reads your file in the browser without uploading it. We also compare the broader options in the best tools to interpret your 23andMe raw data, and if you don't have the file yet, start with how to download your 23andMe raw data.

For the full picture across every category above, see our pillar guide: the complete guide to your 23andMe raw data.

Why coverage is incomplete — and why confirmation matters

It's worth restating the theme that runs under all four categories, because it's the single most important thing to internalise. A genotyping chip is a sampling instrument. It reads a curated subset of positions, so the absence of a finding in your file is weak evidence — the chip may simply never have looked. And the presence of a finding is an association from population data with a real error rate attached, so any single result can be a false positive. Neither direction is conclusive on its own.

That is why confirmation isn't a bureaucratic afterthought — it's the step that turns a raw-data hint into something you can trust. Clinical labs re-test the specific variant with validated methods, a genetic counsellor puts it in the context of your family history and the full set of relevant variants (not just the chip's handful), and only then does it inform a medical decision. Raw data is a good place to raise a question. It is never the place to settle one.

The privacy dimension: your genome is uniquely sensitive

There's a reason we keep coming back to where you read this data, not just what it says. Your genome is the most permanently identifying file you will ever own. You can change a password after a breach; you cannot change your DNA. It also implicates your relatives, who never consented to anything, and it can reveal information — ancestry, health predispositions, family relationships — that you may never have chosen to share. Once a copy of that file sits on someone else's server, you have lost control of it, and the terms under which it's held can change: companies get acquired, policies shift, and what happens to your DNA data when a company is sold is not always in your favour.

This is exactly why Quanome is built the way it is. When you explore your 23andMe or AncestryDNA raw data with Quanome, the file is parsed locally on your own device — it is never uploaded to us. You get the same category-by-category insights described above, read in context alongside your labs and Apple Health data, without your most sensitive file ever leaving your control. For the broader case, see on-device versus cloud health-data privacy. And if you've already uploaded elsewhere and want to pull back, here's how to delete your 23andMe data.

How to explore responsibly

Across all four categories, the same disclaimer holds. Consumer raw data is a screening and curiosity tool, not a clinical one. It carries known false positives and false negatives, it covers only a fraction of your genome, it does not diagnose anything, and a marker is the start of a question rather than the end of one. Explore it out of genuine curiosity, read it privately, treat the low-stakes traits as fun and the serious markers with restraint — and take anything that looks serious to a doctor or genetic counsellor for proper clinical testing. Nothing here is medical advice.

Frequently asked questions

Can 23andMe raw data diagnose a disease?

No. Raw data is for educational and informational use only. It is not clinically validated, it can return false positives, and it does not diagnose anything. Serious markers must be confirmed with clinical testing and discussed with a doctor or genetic counsellor.

What are the most useful things to do with 23andMe data for health?

The most grounded uses are understanding complex disease-risk markers as probabilities, checking carrier status for family planning, reviewing pharmacogenomic context to share with a prescriber, and exploring everyday nutrition and wellness traits — always treating each as a starting point, not a conclusion.

Does a risk marker mean I'll get the condition?

No. For complex, polygenic conditions a marker shifts a statistical probability shaped by many genes plus lifestyle and environment. It is risk, not destiny — many people with a risk variant never develop the condition, and many without it do.

Is it safe to use my 23andMe raw data for health insights?

The data itself is informative, but how you interpret and store it matters. Reading it on your own device, rather than uploading it to a third-party server, keeps the most sensitive file you own from leaving your control.

Does raw data cover my whole genome?

No. Consumer DNA chips genotype only a fraction of your roughly three billion base pairs — typically a few hundred thousand pre-selected positions. Most of your genome is never read at all, so the absence of a variant in your file often just means the chip never looked there.

If a serious variant isn't in my raw data, am I in the clear?

No. A 'variant not detected' result is not reassurance. The chip tests only a small, specific set of positions, so it can miss the variant that matters. For anything serious — hereditary cancer, early dementia, clotting disorders — a negative raw-data result should never substitute for a genetic counsellor and proper clinical testing.

Why do I need to confirm findings with clinical testing?

Consumer raw data is generated on genotyping chips built for scale, not diagnosis. Individual marker readings carry a small but real error rate, so any single result can be a false positive or false negative. Clinical labs re-test the specific variant with validated methods before it informs a medical decision.

What ancestry-linked health context can raw data show?

Some variants are more common in particular ancestral populations — the BRCA founder mutations 23andMe tests are a well-known example. This context can help you and a clinician decide which clinical screening is worth pursuing, but it is background information, not a diagnosis.

Can I act on pharmacogenomic results from my raw data?

Never on your own. Genes that affect drug metabolism are real and clinically important, but consumer raw data is not validated for prescribing. Share the context with your prescriber, who can order proper pharmacogenetic testing if it is warranted — do not start, stop, or change any medication based on a raw-data file.