Science & SocietyKinship & Family Structures

AncestryDNA vs 23andMe for finding relatives: which match list does more

AncestryDNA vs 23andMe for finding relatives: database size, how matches and shared cM are shown, trees, segment tools, privacy and when to test both.

Portrait of Elspeth GilchristElspeth Gilchrist · · 9 min read
AncestryDNA vs 23andMe for finding relatives: which match list does more

For finding relatives, AncestryDNA† usually gives the longer and more useful match list, because its database is roughly twice the size of 23andMe's on the companies' own figures and many of its testers link a family tree to their results. 23andMe earns its place when you want to see where on the chromosomes you and a match overlap, when you want maternal and paternal haplogroups, or when a particular relative you are hoping to find may have tested there and nowhere else. A match can only appear where a relative has tested, so people searching for a birth parent or a lost branch often end up with both.

This comparison draws on each company's help pages and press releases (consulted October 2026), Ancestry's company facts page, Blackstone's 2020 announcement of its purchase of Ancestry, and NPR and STAT reporting from 2025 on 23andMe's bankruptcy and sale. The kits were not tested by us, and details we could not confirm in those sources are left out.

† Links to books and test kits are affiliate links. As an Amazon Associate I earn from qualifying purchases. Links to other shops can earn us a commission as well.

The two match lists side by side

AncestryDNA 23andMe
Database size more than 30 million people in its DNA network, as its company facts page stated in October 2026 about 15 million customers, as cited at a June 2025 congressional hearing (STAT); over 11.5 million research participants, as the company stated in May 2026
Who appears as a match testers who chose, when registering, to see and be seen by matches; changeable later testers who opted in to DNA Relatives; changeable later
How a match is shown shared centimorgans, number of segments and longest segment, placed in a relationship category a predicted relationship with a range, percent DNA shared and number of segments
Smallest match 8 cM (smaller matches removed in 2020) one segment of at least 7 cM and 700 SNPs
Limit on the list none stated; the list loads as you scroll and cannot be exported capped; Premium shows up to three times as many matches
Family trees link your own tree; matches' trees, common ancestors and ThruLines need a family history membership a tree predicted from DNA and ages, showing participants who are third cousins or closer
Shared matches up to 3 without a membership, all with one; enhanced shared matches and clusters with Pro Tools Relatives in Common table; clustering with Premium
Segment tools no chromosome browser segment locations for matches who share them; DNA Comparison of up to 5 relatives with Premium
Haplogroups not among the results its help pages list maternal and paternal haplogroup reports
Raw data download yes, a zipped text file; the email can take up to 24 hours yes; usually about a week, up to 30 days
Uploads from other companies not accepted no upload route described in its help pages

Why database size drives cousin matches

A cousin shows up on your list only if they tested with the same company and switched matching on, and if you share a stretch of DNA long enough to clear the company's threshold. The first depends on database size, the second on how distant the cousin is.

23andMe publishes its own detection figures: about 90% for third cousins, about 45% for fourth cousins and about 15% for fifth cousins. Ancestry's help pages give no equivalent table; its 8 cM threshold is close to 23andMe's 7 cM, so the shape is probably similar.

Now a hypothetical. Imagine 100 living third cousins, 100 fourth cousins and 100 fifth cousins (round numbers for easy arithmetic; in real families the distant cousins far outnumber the near ones). Suppose 8 in 100 of them have tested at one company and 4 in 100 at another. Expected matches are cousins × share tested × chance of detection:

Cousin Chance a tested cousin is detected 8 in 100 tested 4 in 100 tested
Third 0.90 100 × 0.08 × 0.90 = 7.2 100 × 0.04 × 0.90 = 3.6
Fourth 0.45 100 × 0.08 × 0.45 = 3.6 1.8
Fifth 0.15 100 × 0.08 × 0.15 = 1.2 0.6

Halve the share who tested and every row halves. Database size multiplies straight through the list, so a database twice as large tends to give about twice as many matches at every distance. It is also why most of anyone's list is fourth cousins and beyond: detection falls, but the number of cousins grows faster. None of this helps with the one relative you most want. If your half sibling tested only at the smaller company, that is the company that matters.

Reading a match: centimorgans, percentages and a predicted relationship

A centimorgan (cM) is a unit of genetic length; shared cM is the total length of the segments two people have in common. Our cousin calculator works from the coefficient of relationship. A first cousin is related by r = 2 × (1/2)^4 = 1/8, shown as 12.5% and an expected 2 × 1/8 × 3,500 = 875 cM. The worked examples of the coefficient of relationship show where the 1/8 comes from.

Ancestry would place 875 cM in its 1st to 2nd cousin category, which runs from 340 to 1,300 cM and lists first cousin, half first cousin, second cousin, first cousin once removed and others as possible. 23andMe would show a percentage instead: its average for first cousins is 12.5%, and it cites an observed range of 7.31% to 13.8% reported by its customers. Either way, one number fits several relationships, and ages, trees and shared matches must decide between them.

Trees and hints

Ancestry's strength is the trees. ThruLines looks for the same ancestor in your linked tree and your matches' linked trees, then draws possible paths to those matches through common ancestors as far back as your 5th great-grandparents. It needs an Ancestry family history membership, two-step verification and a linked tree that is public or private but searchable. A ThruLines path is a hypothesis built from other people's trees, which can be wrong, so check it against records before you write it into yours.

23andMe builds a tree from the other direction. Its Family Tree predicts how your close DNA Relatives (third cousins or closer) fit together using only DNA and ages, so the ancestors in it start out unnamed and you fill them in. It retired its historical records feature in July 2025, and in March 2026 it added Reconstructed Ancestors, a Premium feature that estimates the ancestry composition of untested parents and grandparents from relatives in your tree.

Segment data and chromosome tools

Ancestry's help page says plainly it has no chromosome browser. You see the number of shared segments and, since 2020, the longest one. Its Chromosome Painter maps your ethnicity regions, not the DNA you share with a match. With Pro Tools, enhanced shared matches show how much DNA your matches share with each other. Ancestry introduced them in June 2024 and named endogamy and pedigree collapse among their uses: where people are related through several lines at once, they share more DNA than any one relationship predicts.

23andMe shows where your segments lie for matches who share them, and its Relatives in Common table can show whether a mutual relative shares DNA with both of you in the same place, a first step toward tying a segment to one ancestral couple. Its DNA Comparison feature, for Premium members, lines up shared segments with up to five relatives at once, measured in cM.

Privacy and ownership

23andMe filed for Chapter 11 bankruptcy on March 23, 2025, in the federal bankruptcy court in eastern Missouri. More than two dozen states sued to halt the sale, and from a June 2025 congressional hearing STAT reported that about 1.9 million of its 15 million DNA testing customers had asked to have their data deleted. The court approved a sale to TTAM Research Institute, a nonprofit led by co-founder Anne Wojcicki, at the end of June, and TTAM completed the purchase on July 14, 2025. Before the sale, TTAM said it would keep the existing privacy policies, honor deletion and research opt-outs, and set up a privacy advisory board. The consumer service still trades as 23andMe, which a March 2026 press release called a nonprofit affiliate of the 23andMe Research Institute.

Ancestry is privately owned. Funds managed by Blackstone completed their purchase of it in December 2020, with GIC keeping a significant minority stake. On both sites, matching is a choice you make, and consent to research is a separate setting. Ancestry also warns that a downloaded copy of your data is outside its security measures.

Widening the net: raw data and uploads

Both companies let you download your raw data as a text file. Ancestry says it does not accept DNA data from other sites. As of October 2026, the main places that take uploads stood like this:

  • FamilyTreeDNA accepts AncestryDNA, 23andMe and MyHeritage autosomal data, with free matching.
  • GEDmatch accepts files from the major testing companies. Of its four privacy settings, "Opt-in" also compares your kit with kits law enforcement submitted to identify perpetrators of violent crimes, so choose deliberately.
  • MyHeritage posted a notice dated August 2025 that it was not accepting DNA uploads.

Policies change, so check each site's current page before uploading.

Which to choose

Searching for a birth parent

Start with AncestryDNA for the size of its list and its trees, then upload the raw data to FamilyTreeDNA and GEDmatch. If no close match appears, test at 23andMe as well. Blaine Bettinger's The Family Tree Guide to DNA Testing and Genetic Genealogy covers testing for adoptees and third-party tools, and our genetic genealogy books by level suggests what to read next.

Distant cousins and brick walls

Ancestry again, for volume and ThruLines, though most fourth cousins and beyond stay unexplained until segments or documents connect them. 23andMe's segment locations, or GEDmatch's free one-to-one comparison, help show that several matches descend from the same couple.

Curiosity about ethnic background

Either kit gives an ancestry estimate. 23andMe adds haplogroups for the direct maternal line and, for testers with a Y chromosome, the direct paternal line. An estimate describes where DNA like yours is found today, not who your people are.

Testing both

The two databases are separate, Ancestry takes no uploads and 23andMe's help pages describe no upload route, so in practice the way into both match lists is a sample to each company. For most people the sensible order is AncestryDNA first, then 23andMe, with a free upload of the Ancestry file to FamilyTreeDNA while you wait for the second result.

Sources and further reading

More in Science & Society