A saliva swab returns a breakdown of a mixed-breed dog's ancestry, sometimes with surprising entries. The method behind it is statistical comparison rather than direct identification.
What is actually measured
The test examines many thousands of positions in the genome where individual dogs are known to differ, rather than reading the whole sequence.
Those positions are chosen because their frequencies vary between breeds, so the pattern across them carries information about ancestry.
No single marker identifies a breed. The result comes from the combination, which is why the analysis is probabilistic throughout.
The reference panel does the work
Each company maintains a database of samples from dogs of verified pedigree, which defines what each breed's marker pattern looks like.
The test dog's pattern is compared against those references, and the algorithm finds the combination of breeds that best accounts for what it sees.
Accuracy therefore depends on the panel. A breed represented by few samples, or absent entirely, cannot be reported correctly.
Why results differ between companies
Different providers use different marker sets, different reference panels and different algorithms, so the same dog can return different breakdowns.
Disagreement is largest for breeds that are genetically similar to each other, and for regional village dog populations that were never formally standardised.
Broad ancestral groupings tend to agree between providers even when the specific breed calls do not, which is a reasonable guide to how much confidence each part deserves.
Why generations back are less reliable
Ancestry becomes harder to resolve the further back it sits, because each generation halves the contribution and the signal weakens.
Recent crosses are called confidently. Small percentages attributed to distant ancestors are often the algorithm accounting for residual variation rather than a real ancestor.
A result listing several breeds at small percentages should be read as uncertainty rather than as a detailed family tree.
The health testing sold alongside it
Many panels also screen for known disease-associated variants, and this is a different kind of test with different reliability.
Carrying a variant does not mean a dog will develop the condition, since many depend on other genes and environment, and some are only established in particular breeds.
A positive result is a starting point for discussion with a veterinarian rather than a diagnosis, and no screening panel substitutes for clinical assessment of a dog showing signs.