AncestryDNA Raw Data: Download and Analyze It (2026)

The File Ancestry Gives You Is Bigger Than the Report You Read

TL;DR: Your AncestryDNA test genotyped roughly 700,000 positions in your DNA. Ancestry uses about 300,000 of them to estimate your ethnicity regions, and interprets none of them for health — AncestryHealth was discontinued in 2021. You can download the raw file from DNA Settings → Download or Delete → Download DNA data (the email can take up to 24 hours, and the link works once). That file is a plain tab-separated text file you own, and it can be re-analyzed for pharmacogenomics, nutrition-related variants, carrier status, and polygenic risk scores — the layers Ancestry never computed, each with real limits an array cannot escape.

If you took an AncestryDNA test, you got a map with coloured regions, a list of DNA matches, and maybe a few trait results about freckles or cilantro. Then the novelty wore off.

What most people never realize is that the ethnicity estimate is a summary, not the data. Underneath it sits a file with hundreds of thousands of individual genotype calls — and Ancestry reads only a fraction of them, for one narrow purpose.

It is a letter addressed to you that was opened, skimmed for one paragraph, and filed away. The rest of the page is still there.

This guide covers what is in that file, how to download it, how to read its format (which differs from 23andMe's in ways that break naive parsers), what Ancestry does not tell you, and what else the same data supports. Tested with 23andMe instead? We have a separate guide.

How Many SNPs Are in an AncestryDNA File?

Ancestry's current documentation states that the test examines "approximately 700,000 individual locations of DNA (SNPs) on chromosomes 1-22 and the X chromosome," and that approximately 300,000 of those are used to infer ancestral regions.

That gap is the whole point of this article. Roughly 400,000 genotyped positions play no part in the ethnicity estimate you were shown, and none of your file is interpreted for health. (Relative matching does draw on the dense autosomal set, not just the ~300,000 used for regions.) You can check which array you got by opening the file and reading its header.

Chip Versions Matter

AncestryDNA has used two genotyping arrays, and which one you got depends on when you tested:

The two versions share only about 380,000 positions — roughly 56% of V1 or 60% of V2, depending which you use as the denominator. Ancestry has never published exact marker counts or an official overlap figure; these come from measured raw files catalogued by the genetic-genealogy community, cross-checked against Ancestry's raw data documentation (verified August 2026). This matters if you are comparing your file against a relative's from a different era, or against a study that assumed a specific marker set: a variant present in one file can be simply absent from the other. Absent is not the same as negative — a position that was never genotyped tells you nothing at all.

All AncestryDNA positions are reported on genome build GRCh37, forward strand. The UCSC equivalent is hg19, though the two use different mitochondrial reference sequences — worth knowing if you touch the chromosome-26 rows. Any tool you use must assume the same build, or every position will be silently misaligned.

How to Download Your AncestryDNA Raw Data

Ancestry still offers raw data download in 2026, and no insolvency event puts the feature in question — Blackstone acquired the company in December 2020 and it operates normally.

Step by Step

1. Sign in and open DNA Settings for the test you want.
2. Find the Download or Delete section.
3. Choose Download DNA data.
4. Tick the confirmation checkbox and select Continue.
5. Re-verify your identity — either enter your current password, or have a code sent to your email and enter it.
6. Wait for the confirmation email, then click Download DNA Data in it.

Three practical details Ancestry documents that catch people out:

• The email can take up to 24 hours. Often much faster, but if nothing arrives in ten minutes that is normal, not a failure.
• The link is single-use and expires after one week. Save it to a device you actually keep, not a borrowed laptop.
• If you request it twice, only the most recent email works. Older links go dead.

You will receive a .txt file inside a ZIP named dna-data-(date).zip.

One asymmetry worth knowing: Ancestry lets you take your data out but not bring other data in. Your file is portable away from Ancestry, not toward it.

Understanding the AncestryDNA File Format

Open the file in any text editor. It begins with comment lines starting with #, including the generation date and the array version, plus a notice in capitals that the data is not intended for medical, diagnostic, or health purposes.

After the comments comes a header row and then your genotypes, as five tab-separated columns:

rsid	chromosome	position	allele1	allele2
rs4477212	1	82154	A	A
rs3094315	1	752566	G	G
rs2073813	1	753541	0	0

Each row is one position: its reference SNP ID, its chromosome, its coordinate on build 37, and your two inherited alleles in separate columns.

The Details That Break Parsers

Four conventions differ from what most tools expect, and each one causes a specific class of silent error:

• Chromosomes are numbers, not letters. 23 is X, 24 is Y, 26 is mitochondrial DNA, and 25 is the pseudoautosomal region of the X. Ancestry documents only the 23/24 mapping; 25 and 26 are convention across parsing tools. Its "1-22 and X" headline figure ignores the last two — a V2 file also carries 1,691 Y and 195 mitochondrial markers.
• No-calls are zeros. A position the array could not read appears as 0 in both allele columns. A parser that treats 0 as data will happily record a genotype of "00" for a position that was never measured.
• Insertions and deletions use I and D rather than sequence.
• Alleles are in two columns, not one. Tools written for 23andMe expect a single fused genotype field.

Here is the same information side by side:

This is why "upload your raw data" tools sometimes report a suspiciously low variant count, or a strange number of homozygous calls, when fed an Ancestry file. The format is not harder, just different — a parser has to be told which one it is reading.

What AncestryDNA Does Not Tell You

This is the part that surprises people most, and it is not a criticism of Ancestry — it is a deliberate scope decision the company made and announced.

AncestryHealth no longer exists. Launched in 2019, discontinued in January 2021 with support continuing through mid-2021, when Ancestry announced it was deepening its focus on family history. The product exists in no market today, Europe included.

What remains adjacent to health is AncestryDNA Traits, and Ancestry is explicit in its own documentation that Traits "is not designed or intended to diagnose or provide medical information related to health conditions" and "is not a substitute for professional medical advice or diagnosis." The catalogue reflects that: eye colour, freckles, cilantro aversion, earwax type, and similar. These are genuinely interesting and genuinely cosmetic.

So the situation is specific: you have roughly 700,000 genotyped positions, an ethnicity estimate built from about 300,000 of them, a match list, and a set of cosmetic traits. Nothing has been read for pharmacogenomics, carrier status, or clinically annotated variants — that was never what you bought.

The genotypes are sitting there regardless. Re-analysis is interpretation, not new lab work.

What the Same File Can Support

Below is what a raw genotype file of this density can and cannot support. The distinction matters more than the list.

Pharmacogenomics — How You Process Medications

Some of the best-established findings in consumer genomics are pharmacogenomic. Variants in CYP2C19 and VKORC1 change how fast you metabolize or respond to specific drug classes, and the common alleles are usually present on consumer arrays. Clinical guidelines exist for a number of drug-gene pairs, which is unusual in this field and makes them comparatively solid ground.

SLCO1B1 works differently, and the distinction matters: it encodes the OATP1B1 uptake transporter, so a reduced-function variant raises statin exposure by slowing hepatic uptake — not by slowing metabolism.

CYP2D6 is the important exception, and it is worth being blunt about it. It is one of the most clinically relevant pharmacogenes and one of the hardest for an array to read, because a significant minority of people carry gene deletions, duplications, or CYP2D6/CYP2D7 hybrids that a SNP chip cannot see. An array can report the star-allele markers it genotyped; it cannot establish your CYP2D6 copy number. Our CYP2D6 gene page goes into why, and when a clinical-grade test is the right next step instead.

This is information your prescriber acts on, not you. It is a conversation to have with a clinician, not a reason to change a dose. Our guide to pharmacogenomics in Europe covers how this works in practice, and you can read about individual genes on our gene reference pages.

Nutrition-Related Variants

Lactase persistence has a genuinely clean story: a single regulatory variant upstream of LCT keeps the lactase gene switched on into adulthood. Caffeine metabolism, folate processing and vitamin D binding are more contested — the biochemistry is real, but the leap from genotype to a change in your diet is where the overselling happens.

MTHFR is the clearest case of that gap: the American College of Medical Genetics advises against testing it in routine clinical evaluation, which has not stopped the wellness industry from building a cottage industry on it. See our nutrigenomics guide for where the evidence holds and where it thins out.

Carrier Status — With an Important Limitation

Array genotyping checks specific known positions; it does not sequence a gene. Where a handful of common variants account for most carriers, an array is useful. Where pathogenic variants are spread across hundreds of rare positions, it can miss them entirely — and "nothing found" is not a clear result. It means the positions tested were negative.

The reverse error matters even more, and it is the one nobody mentions. SNP chips are excellent for common variants — the same UK Biobank analysis puts sensitivity, specificity and predictive value all above 99% there. But for variants rarer than about 1 in 100,000 the positive predictive value collapses: only 16% of rare pathogenic heterozygous calls survived sequencing confirmation (Weedon et al., BMJ 2021). A "pathogenic variant found" flag derived from array data is more likely to be an artefact than a finding.

That is a reason to seek clinical-grade confirmation, never a reason to act. Anyone with a family history that concerns them should take that to a doctor or genetic counsellor, who can arrange proper testing — a re-read of a consumer file is not a substitute.

Polygenic Risk Scores

Aggregating many small-effect variants produces a score that is meaningful across a population and modest for an individual. Most published scores were developed in cohorts of largely European ancestry, which limits how well they transfer to everyone else. We wrote about what polygenic risk scores can and cannot do because the gap between the two is where most misunderstanding lives.

What an Array Cannot Do

It cannot find variants at positions it does not test, and it cannot detect structural variants or repeat expansions at all from the downloadable file, which holds genotype calls but none of the probe-intensity data that copy-number inference needs. It is not a substitute for sequencing. If you want to understand why, our explainer on what SNPs are covers the mechanics, and the whole genome vs. exome comparison covers what sequencing adds.

Privacy: What Changes the Moment You Download

Ancestry states it plainly in its own download documentation: once you download your DNA data, that copy is no longer under the protection of Ancestry's security measures. That is a real transfer of responsibility to you, and the strongest argument for being deliberate about where the file goes next.

A few facts to weigh, all from Ancestry's own published material:

• European users are served by Ancestry Ireland Unlimited Company in Dublin, which acts as data controller for users outside the United States. Transfers to the US rely on standard contractual clauses and the Data Privacy Framework.
• Ownership sits with Blackstone, which completed its acquisition in December 2020 at a $4.7 billion enterprise value, with GIC retaining a significant minority stake. It is a US-headquartered company, which is the fact that matters for the point below.
• Law enforcement: Ancestry's transparency report published in February 2026 reports zero valid requests for access to customers' genetic database in the period covered, and nine valid requests for non-genetic data, of which seven were fulfilled. In December 2025 Ancestry restated that law enforcement cannot use its DNA tools, trees, or records to investigate crimes or identify remains.

None of that is a reason for alarm, and none of it is a reason for complacency. It is the actual privacy posture, which is more useful than a vague sense of unease. Your rights over this data in Europe are concrete and enforceable — we covered them in GDPR and genetic data.

The structural point survives any change of management: a company incorporated outside the EU answers to a legal system you did not vote for, however good its current intentions. That is not a claim about Ancestry's conduct — by its own transparency reporting, that looks careful. It is a claim about where the decision sits. Genomic data is the one dataset you cannot rotate like a password, so jurisdiction over it matters more here than almost anywhere else.

When you choose where to upload your file, the questions worth asking are where it is stored, whether it is encrypted at rest, whether it is used to train anything, and whether you can delete it and have that mean something.

Frequently Asked Questions

Can I still download my AncestryDNA raw data in 2026?

Yes. The download lives under DNA Settings → Download or Delete → Download DNA data. You re-verify your identity with either your password or an emailed code, then receive an email with the download link. Ancestry notes the email may take up to 24 hours, and the link is single-use and expires after one week.

How many SNPs does an AncestryDNA file contain?

Around 700,000 positions. The V2 array used since May 2016 reports 668,942 total and 637,639 autosomal SNPs. The earlier V1 array reported roughly 682,000. The two versions overlap by only about 60%.

What genome build does AncestryDNA use?

GRCh37, also called hg19, on the forward strand. Any analysis tool must assume the same build or the coordinates will not line up.

Why does my AncestryDNA file look different from a 23andMe file?

Because it is a different format. AncestryDNA uses five tab-separated columns with the two alleles in separate columns, numeric chromosome labels where 23 is X and 26 is mitochondrial, and 0 for positions that could not be read. 23andMe uses four columns with the alleles fused into one field, letter labels for X, Y, and MT, and -- for no-calls.

Does AncestryDNA give health results?

No. AncestryHealth was discontinued in January 2021 and does not exist in any market. AncestryDNA Traits covers cosmetic and behavioural characteristics, and Ancestry's own documentation states it is not intended to diagnose or provide medical information.

Can I upload my AncestryDNA file somewhere else?

Yes — the file is yours and portable. Note the asymmetry: Ancestry does not accept DNA data from other services, so the portability runs outward only.

Is a raw data re-analysis a medical test?

No. Consumer array data is not a diagnostic instrument, and no re-interpretation of it changes that. It can be genuinely informative, particularly for pharmacogenomics, and it can raise questions worth taking to a clinician. It cannot answer them.

What to Do Next

If your AncestryDNA file is still sitting in your account, download it — the link expires, ownership changes happen, and a file on your own drive is a file you control.

Then decide whether the ~400,000 positions Ancestry never interprets for anything beyond relative matching are worth reading. For pharmacogenomics in particular, the answer is often yes, because the evidence base is unusually solid and the findings are the kind a clinician can actually use.

DeepDNA analyzes AncestryDNA files (V1 and V2) alongside 23andMe and MyHeritage, auto-detecting the format from the file header and normalizing chromosome labels and no-calls before anything is interpreted. Every position is handled on GRCh37 — the build Ancestry reports — so no coordinate conversion is involved. Reports are generated in the EU, the data is encrypted at rest, and it is a one-time payment of €29 for the analysis and report — no subscription, and nothing that renews.

You already paid for the genotyping. The interpretation is the part that was missing — and the file is yours either way, whoever you decide should read it.

This article was created with AI assistance and reviewed by the DeepDNA editorial team.