How DNA Matching Works: From SNP Array to Match List

The match list inside an AncestryDNA, 23andMe, or MyHeritage account looks like magic: thousands of named strangers sorted by how closely related they appear to be. The mechanism behind it is the more useful part of the result, and it is the part most customers do not look at. Once you know how matching works, the cM numbers stop being mysterious and start being a practical tool.

What a consumer DNA test actually reads

A consumer ancestry test does not sequence your whole genome. It runs your saliva sample through a genotyping chip that reads several hundred thousand specific SNPs, the positions in the genome that vary between people. The test result is a file of letters at each tested position, your genotype. This is the same input every step of the result uses, including the ethnicity estimate and the match list.

How two people are compared

To identify matches, the company compares your genotype to every other genotype in its database, looking for stretches of chromosome where you and another person share the same letters at consecutive SNPs. A long enough shared stretch is unlikely to occur by chance, and is almost always evidence that you and the other person inherited that segment from a common ancestor. Geneticists call this “identical by descent,” or IBD.

The unit used to measure segment length is the centimorgan, abbreviated cM. A centimorgan is a unit of genetic distance, not physical distance, and it reflects the probability of recombination occurring across that segment in a single generation. Two people share a certain number of total cM across all their shared segments, and a longest shared segment of a certain length. Our understanding centimorgans explainer covers the unit in depth.

What total cM and longest segment tell you

Total shared cM is the strongest signal of how closely you and another person are related. Full siblings share around 2,500 cM total on average, first cousins around 850 cM, second cousins around 230 cM, third cousins around 75 cM, and the numbers fade quickly past that. The shared DNA and relationship chart lays out the full table from the Shared cM Project.

Longest segment matters because it discriminates between recent and distant shared ancestry. Two people who share 100 cM total in one 100 cM segment are almost certainly closer relatives than two people who share 100 cM total spread across ten 10 cM segments, because longer segments survive fewer generations of recombination.

Match lists are sorted by total shared cM

Every major consumer database sorts your match list by how much DNA you share. The person at the top, after parents and siblings if any have tested, is the person who shares the most cM with you. The list trails off into dozens and then thousands of more distant matches.

The companies usually hide very small matches by default, often below around 6-8 cM, because below that threshold the signal-to-noise ratio deteriorates and matches there are more likely to be false positives than real distant cousins. We do not recommend turning that cutoff off.

Each company’s database is separate

A match list at AncestryDNA only contains people who tested at AncestryDNA. A match list at 23andMe only contains people who tested at 23andMe and opted into DNA Relatives. This is the most important practical fact about DNA matching, and it surprises customers constantly. If your biological half-brother tested at AncestryDNA and you tested at 23andMe, you will not see each other in either system.

There are two workarounds. First, several companies accept free raw data uploads. MyHeritage, FamilyTreeDNA, and the third-party site GEDmatch all accept uploads from AncestryDNA and 23andMe. Uploading puts your data into more matching databases without buying a second test. Second, the third-party site GEDmatch lets you compare against people from any source company who have also uploaded there.

AncestryDNA has the largest database, at roughly 25-30 million profiles as of early 2026, which is why it usually returns the most matches. 23andMe holds roughly 14 million profiles post-bankruptcy, with the corporate situation still in flux; see our genetic data privacy guide for the full context.

What matching cannot tell you

Shared cM tells you a relationship range, not a single relationship. Two people who share 1,750 cM could be a parent and child, or full siblings, or in unusual cases a grandparent and grandchild. The math constrains the options to a small set; it does not pick one.

Matching also does not tell you which side of your family a match is on without additional information. Companies that offer “parental side clustering” tools, such as 23andMe and AncestryDNA’s SideView, use match networks and triangulation to infer maternal versus paternal sides, but that inference is not perfect.

For how to use matches to build out a family tree, see building a family tree from DNA and the ancestry pillar guide.