The raw data download is a plain text file: one line per position the chip reads, four columns per line. This guide shows a real example and explains every part of it.
Written by Martin Pechhacker, Vienna Last checked 18 September 2026 10 min read
Written by Martin Pechhacker, Vienna
Last checked 18 September 2026
10 min read
The download is a zip archive whose name starts with genome_. Inside it is one plain text file with tab-separated columns.
Every line is one position: an identifier (rsid), the chromosome, the position on reference build 37 (GRCh37) and your genotype, usually as two letters.
A file from the current chip (v5) has about 640,000 lines. That is a list of read positions, not your genome: the sequence between them is not in the file.
-- means the chip could not read that position. Single letters appear on the X and Y chromosomes of men and on the mitochondrial DNA of everyone.
Below is the top of a 23andMe raw data file: the comment lines 23andMe writes at the start, the column line, and three data rows. The comment lines are quoted from a file that a Personal Genome Project participant published under open consent; the three rows are the sample genome that runs the demo, written in the same layout. Sources 2 4
Sources 2 4
# This data file generated by 23andMe at: Tue Jun 12 21:48:40 2018
# This file contains raw genotype data, including data that is not used in 23andMe reports.
# Below is a text version of your data. Fields are TAB-separated
# Each line corresponds to a single SNP. For each SNP, we provide its identifier
# (an rsid or an internal id), its location on the reference human genome, and the
# genotype call oriented with respect to the plus strand on the human reference sequence.
# We are using reference human assembly build 37 (also known as Annotation Release 104).
# rsid chromosome position genotype
rs4988235 2 136608646 AA
rs713598 7 141673345 CC
rs1815739 11 66328095 CC
The first lines of a 23andMe file. Two runs of comment lines are left out, marked [...]: they describe the research-only status of the data and link to 23andMe's help pages. The three rows are the demo genome's calls at three well-known positions. Sources 2 4
Everything after the column line follows the same pattern, one position per line, in chromosome order: chromosome 1 first, then 2 to 22, then X, Y and MT. The four fields are separated by a tab character.
One line of the file, field by field.
rsid
The position's public name in dbSNP, the reference list of known variants, so the same name means the same position in every database and paper. Some lines carry an i number instead: an identifier 23andMe assigned to a probe of its own, which other databases may not know.
chromosome
1 to 22, X, Y, or MT for the mitochondrial DNA. The Y lines are present in every file; in a file from a woman they all read --.
position
The base number on the reference genome, counted from the start of the chromosome. 23andMe uses build 37 (GRCh37), and the header says so. The same place has a different number on build 38, which is why a reader must know which build a file uses before it does anything with it.
genotype
Your two alleles at this position, one inherited from each parent, as the letters A, C, G and T. AA means both copies carry A; AG means one copy carries A and the other G. The letters are given on the plus strand of the reference genome, so they can be compared with public databases directly.
That is the whole format. A reader takes each line, looks the rsid up in the research it holds, and reports what your two letters mean there. The screenshot shows the result for the second example row, rs713598, a position in the TAS2R38 gene that shapes how bitter some foods taste.
The card for rs713598 on the demo genome: one line of the file, read. Source 4 Open this card in the demo
23andMe delivers the raw data as a zip archive. Its name starts with genome_, followed by the name on the account, the word Full and a timestamp, and files from the current chip carry v5 in the name as well. Inside is one text file with the same name and the ending .txt, about 15 MB for a v5 file. Upload the zip as it is; there is no need to unzip it. Source 5
Sign in to 23andMe.
Open your profile icon (top right) and choose "Settings".
Scroll to "23andMe Data" and press "View" (it may ask for your date of birth).
Under "Raw Data", press "Request your raw data" (or "Submit request").
When the email arrives, follow its link, or come back to the same page, and press "Download Raw Data".
Preparation usually takes a few hours and can take one to two days. 23andMe emails you when the file is ready. The file is prepared from the genotypes 23andMe holds for you, so requesting it changes nothing on your account.
READING THE GENOTYPE COLUMN
Most lines carry two letters, but not all. The other forms each mean something specific, and a reader has to handle every one of them.
WHAT YOU SEE
WHAT IT MEANS
AA, AG, CT and so on
Two alleles read, one per copy of the chromosome. The order of the two letters carries no meaning.
--
No call: the probe for this position gave no clear signal in your sample. Some thousands per file are normal, and which positions are affected differs from sample to sample.
A, C, G or T alone
One allele, because there is only one copy to read: the X and Y chromosomes in men, and the mitochondrial DNA in everyone. A man's file shows single letters on X everywhere except the short stretches the X and Y share.
D, I, DD, DI, II
A small insertion or deletion rather than a single-letter difference: D for the deletion, I for the insertion. A few thousand positions per file are written this way.
i before the number
A 23andMe internal identifier in place of an rsid, used for probes 23andMe designed itself. About 18,000 of the 640,000 lines in a v5 file.
Counted in the two public files behind this guide: a v5 file from 2018 and a v3 file from 2011. Sources 2 3
Sources 2 3
A line is a read, not a certainty
The header of every file says that the data has undergone a general quality review and that only a subset of markers have been individually checked. A single line can be wrong, and the same person genotyped twice gets a small number of different calls. Any report built on the file inherits that, which is why a careful reader shows how well a position is usually read alongside what it found there.
23andMe has used five genotyping chips since 2007. The chip decides which positions are in the file and how many lines it has, so two files from different years are not the same list of positions. The file does not name its chip, but the line count and the date in the first header line usually settle it. Source 1
CHIP
IN USE FROM
LINES IN THE FILE
v1
November 2007
Rare today; none among the files ReadYourDNA has processed
v2
September 2008
About 555,000
v3
November 2010
About 960,000
v4
November 2013
About 577,000
v5
August 2017
About 610,000 to 640,000
Dates from the chip-version table the genetic-genealogy community maintains; line counts as seen in the files ReadYourDNA has processed, and for v2 from the same table. Source 1
The lists overlap heavily but are not identical, so a position one chip reads may be missing from another. A report built on the file can only cover what the chip read, and a good report says so for each of its sections rather than filling the gap.
WHAT A READER CAN DO WITH IT
Every 23andMe file, whatever its chip, carries the four columns above, so each of the four ReadYourDNA reports can be built from it. What changes with the chip is coverage: how many of the positions a report needs are in the file. Carrier annotation feels it most, because it looks for specific rare positions that a chip either includes or does not; polygenic scores feel it least, because they add up thousands of common positions and a missing few change little. Source 6
REPORT
WHAT IT NEEDS FROM THE FILE
WHERE THE CHIP MATTERS
Variant explorer
The rsid and genotype of every position it holds a card for
A card appears only for positions in your file. A v4 file has fewer than a v5 file, and each card says whether the position was read.
Polygenic trait scores
Thousands of common positions per trait
Least affected: the score models tolerate missing positions, and every trait page states how many of its positions your file covered.
Medication response
The star-allele positions of 23 pharmacogenes
Most of them are on every chip. A gene whose positions are missing is reported as not called rather than guessed.
Carrier annotation
Specific rare positions from ClinVar
Most affected. In the files ReadYourDNA has processed, a v5 file reads positions for about half of the 3,321 conditions in the catalog, a v3 file about 40 percent, a v4 file about 14 percent. The report shows the number for your file.
Coverage as measured in the files ReadYourDNA has processed; every report states the figure for your own file. Sources 6 7
Sources 6 7
ReadYourDNA reads a raw DNA file you already have and returns polygenic trait scores, medication-response annotation, carrier annotation and an annotated variant explorer, with the study behind every number. One payment, no subscription, and you can delete the file at any time.
Educational, not medical advice.
The sequence between the read positions. A v5 file lists about 640,000 positions out of roughly 3 billion in the genome. Everything between them is unread, so a variant that is not on the chip is simply absent, and its absence says nothing about whether you carry it.
Larger changes. The chip reads single positions and a few thousand small insertions and deletions. Duplicated or missing stretches of a chromosome, repeat expansions and rearrangements are not in the file.
Anything about how genes are used. The file has no information on gene activity, methylation or anything else that changes over a lifetime; it records inherited letters only.
Your 23andMe reports. Ancestry composition, relative matches and the reports on the 23andMe site are computed from the file but are not in it. The download is the input, not the output.
Your name and details. Beyond the account name in the file name, the file holds no identifying fields. The genotypes themselves are unique to you and, for the most part, to an identical twin.
Yes. It is plain text with tab-separated columns, and Excel, Numbers and LibreOffice all open it. With over 600,000 rows it is slow to load, and a spreadsheet may reformat a column when it saves, so keep the original file for uploads and open a copy to look around. A text editor shows it exactly as it is.
Build 37, and the header says so on every file 23andMe currently produces. Older files, generated up to about 2012, say build 36 instead. A reader that expects build 38 has to convert the positions first; ReadYourDNA reads the build from the header and converts internally.
It can. The header itself says that data downloaded at different times may differ because of ongoing improvements in genotype calling. The list of positions stays the same; a small number of calls may change, and a no-call may become a call.
Count the lines, or look at the date in the first line of the header. About 640,000 lines and a date after August 2017 means v5; about 577,000 lines means v4; about 960,000 lines means v3. After an upload, the ReadYourDNA carrier status report names the chip it recognised.
Each position is read by a probe on the chip, and a probe sometimes fails to give a clear signal for a sample. Those positions are written as --, the no-call. Some thousands per file are normal and are not a fault in your sample.
No. A whole-genome sequence reads every position and is delivered as a VCF file, usually hundreds of megabytes; a 23andMe file is a fixed list of a few hundred thousand positions read by a chip. Both can be uploaded to ReadYourDNA, and the methodology page explains how each is handled.
Where each fact came from
Every figure on this page was read from the page named, on the date named. ReadYourDNA publishes this page. If something here is out of date, write to support@readyourdna.com.
Roots & Recombinant DNA, "23andMe Chip Versions Comparison"
The five chip versions and the month each entered use, with the v2 marker count, citing the ISOGG wiki's chip table.
Read on 18 September 2026
Personal Genome Project participant hu005023, public 23andMe data file (generated 12 June 2018)
The header lines quoted in the example, the column layout, the line count (638,468) and the counts of no-calls, insertion and deletion codes and internal identifiers.
Fields are TAB-separated
We are using reference human assembly build 37 (also known as Annotation Release 104).
Note that it is possible that data downloaded at different times may be different due to ongoing improvements in our ability to call genotypes.
Personal Genome Project participant hu002B3C, public 23andMe data file (generated 13 May 2011)
A build 36 file from a man: 966,983 lines, single-letter calls on X, Y and MT.
We are using reference human assembly build 36.
The three example rows and the variant card screenshot, on the sample genome.
Which files are accepted, and that the zip can be uploaded as it is.
How array files and sequencing files are read, and what each report needs.
ReadYourDNA carrier status
The catalog size and the per-file coverage line.
ReadYourDNA and Promethease
ReadYourDNA · About ReadYourDNA · FAQ · Methodology & Science · Pricing · Support and complaints · Legal & Contact · Health Traits · Pharmacogenomics · Pharmacogenes · Carrier Status · Variant Explorer · The Human Karyotype · Privacy Policy · Terms of Service · Accessibility · Consumer Health Data Privacy · Your Privacy Choices · ReadYourDNA and Promethease compared · What is inside a 23andMe raw data file
Written, maintained and operated by Martin Pechhacker, sole proprietor, Vienna, Austria. Trade register GISA 40064420. support@readyourdna.com