Start with the paperwork — adoption record, first vet chart, microchip registration, breeder papers. With none of that, describe the dog by structure rather than colour and narrow to a breed group, not a breed. Eyeballing has a ceiling: for 14 of 20 dogs of known ancestry, fewer than half of 923 professionals named any breed the DNA held.
‘What breed is my dog’ is two questions wearing one coat. One is about ancestry, which is a genetics question with a factual answer. The other is about appearance, which is a description. For a registered purebred the two agree and the paperwork settles it in a minute. For a mixed dog they come apart routinely, and how far apart has been measured rather than guessed.
The measurement is uncomfortable. In the 2013 study by Voith and colleagues, 923 people who work with dogs professionally were shown 20 dogs whose DNA ancestry was known. For 14 of the 20 dogs, fewer than half of those professionals named any breed that appeared in the DNA result. For only 7 of the 20 could even half the observers agree on a predominant breed — and in 3 of those 7, the breed the majority agreed on was not in the DNA result at all, as a major or a minor component. Confidence and correctness turned out to be separate things.
The label also explains less than people hope. In 2022 Morrill and colleagues surveyed the owners of 18,385 purebred and mixed-breed dogs and genotyped 2,155 of them, then asked how much of the behavioural variation between individual dogs breed accounts for. The answer was about 9%. Whatever name you land on, it is a weak predictor of the animal in front of you.
The adoption or rescue record, the first vet chart (vets usually write down a breed impression at the first visit), the microchip registration entry, any breeder or kennel-club papers, and the original listing if the dog was bought online. If the dog is registered with a recognised kennel club, the pedigree is the answer and nothing else is needed. This step costs nothing and it is the only step that can produce documentary evidence instead of an opinion.
Write down adult height and weight, head shape and how long the muzzle is compared with the skull, ear set (drop, prick, rose, button), where the tail is set and how the dog carries it, coat type (single or double, wire, curly, smooth), foot shape, and whether the topline is level or sloping. Then search on those. Colour is the trait your eye grabs first and the one that most often sends people to the wrong breed, because unrelated breeds share the same colours and coat patterns while sharing nothing else.
Herding dogs, retrievers and spaniels, scent hounds, sighthounds, terriers, northern spitz types, livestock guardians and mastiffs — each group shares a body plan because each was shaped around a job. ‘A sighthound-shaped dog’ is a claim you can defend from a deep chest, a tucked waist and long legs. ‘A whippet’ is a claim you almost certainly cannot. Stopping at the group is not giving up; it is the level at which a visual answer is still honest.
Take three photos — standing side-on, head-on, and a few seconds of video of the dog trotting — and show the same set separately to your vet, a groomer and a trainer. Write down all three answers before anyone hears anyone else's. If they match, your visual read is about as good as visual reads get. If they do not, you have just reproduced the Voith result on a sample of three, and you now know what a single confident answer is worth.
A cheek-swab DNA kit is the only method that returns an ancestry breakdown rather than an impression. Embark's own instructions are to keep the dog from eating for 30 minutes beforehand and to swab for 30 to 60 seconds, aiming for the full 60, with results in two to four weeks for its breed tests. Do it when the answer changes something concrete: a landlord or insurer with breed restrictions, a health screening decision, or a suspected breed-linked condition your vet wants to rule in or out. Ask the vet which health findings would actually change your dog's care before you pay for a health panel on top of the breed one.
This page was written by DKsequence. We also make the Android app recommended below — the steps above work whether or not you install it.
A photo tool is doing visual identification, which is the method that came apart in the Voith study when 923 professionals tried it on 20 dogs of known ancestry. It can be a useful way to generate a shortlist. Treat the output as a hypothesis to check against structure and, if it matters, against a DNA test. Woofpedia is not one of these tools — it has no camera and takes no photos.
Embark, as one published example, says to keep the dog from eating for 30 minutes before swabbing, to swab for 30 to 60 seconds, and to expect results in two to four weeks for its breed tests. The wait matters for planning: if you need a breed written on a form this week, the test will not arrive in time and you will be back to describing the dog.
It is usually a guess made by eye, in seconds, by someone looking at dozens of dogs that week. That is close to the setup the Voith study tested, and in it a majority of professionals agreed on a predominant breed for only 7 of 20 dogs, and were wrong about that breed for 3 of those 7. Read a shelter label as a description of how the dog looked to one person on one day.
Barely. The 2022 study of 18,385 surveyed dogs and 2,155 genotyped ones put breed at roughly 9% of the behavioural variation between individual dogs. Breed shifts the odds on a few specific tendencies; it does not describe the animal on your sofa. A month of watching your own dog will outperform the label every time.