Jewish ancestry DNA test: what ancient DNA and a qpAdm model can tell you that a percentage cannot

If you have Jewish ancestry and have taken a consumer DNA test, you have probably seen a single line in your results: "Ashkenazi Jewish", "Sephardic and North African Jewish", or a similar label with a percentage beside it. That number is not wrong, but it answers a narrower question than most people think it does. It says how closely your genome resembles a reference group of present-day people who identify as Jewish. It does not say where that ancestry came from, how much of it is Levantine, how much is southern European, Iranian, Arabian or Caucasian, or whether any of those claims would survive a statistical test.

This guide is about the second set of questions, which is where ancient DNA and the qpAdm method come in. It is written from the inside, since we sell Ancient Ancestry, a qpAdm analysis, so it tries to be exact about what the method can and cannot say for a Jewish genome rather than persuasive. It sets out what ancient genomes have established, then why a qpAdm model rather than a percentage tool: what each one measures, and a checklist to take to any seller. That part is for you if you already have a raw DNA file from 23andMe, AncestryDNA, MyHeritage or FamilyTreeDNA and are deciding whether a model would tell you more than the percentages. The rest covers what a report shows, the bar every published model clears, and what no model can do.

The companion posts go deeper on each community: Ashkenazi, Sephardic, Mizrahi, Yemenite and the Mountain, Georgian and Bukharian communities, plus the Canaanite source population that all of them share.

One thing first: ancestry is not identity#

A genome is not a membership card. Nothing in this guide, and nothing in any report we publish, is a statement about whether someone is Jewish, about halachic status, about conversion, or about belonging. Jewish identity is religious, cultural, legal and familial, and none of those things are measured by allele frequencies. A convert's genome may carry no Levantine ancestry at all, and the same Levantine ancestry is carried by people of many religions, languages and countries. A qpAdm model describes a genome's ancestry relative to sampled ancient populations. That is the whole claim, and it is a claim about biology, not about who you are.

What ancient DNA has established#

Two decades of population genetics on living Jewish communities, and a much shorter but decisive period of work on ancient genomes, agree on a picture that is now fairly stable.

Most Jewish communities share a Levantine core. The genome-wide studies of 2010 showed that Ashkenazi, Sephardic, Italian, Syrian, Iraqi, Iranian and Kurdish Jewish communities cluster together and with the non-Jewish populations of the Levant, rather than each community clustering with its host population (Behar et al. 2010; Atzmon et al. 2010). The exceptions are the Ethiopian and Indian communities, whose genomes largely resemble their neighbours, and which are best understood as communities with a different demographic history.

That core is Bronze Age Canaanite. Ancient genomes from Megiddo, Hazor, Sidon, Abel Beth Maacah and other sites of the second millennium BC describe a population that was a blend of the local Levantine Neolithic (itself descended from the Natufian foragers) with a substantial Iranian- or Caucasus-related component that had been arriving since the Chalcolithic (Haber et al. 2017; Agranat-Tamir et al. 2020). When present-day Jewish and Arabic-speaking Levantine groups are modelled against these genomes, a Canaanite-related population supplies a large part of their ancestry, with the remainder differing by community.

The remainder is where the communities diverge. Ashkenazi genomes carry a southern European component, mostly Italian-like, that amounts to roughly half of their ancestry by most estimates, plus a smaller eastern European share (Xue et al. 2017; Waldman et al. 2022). Sephardic communities of Turkey, the Balkans and North Africa carry Iberian and local admixture on the same base. Iraqi, Iranian and Kurdish Jews carry more Iranian-related ancestry and less European. Yemenite Jews sit between the Levant and Arabia. The Caucasus and Central Asian communities add local Caucasian and Central Asian ancestry to a Persian-Jewish core.

Founder effects shape everything. The Ashkenazi population passed through a severe bottleneck around 600 to 800 years ago, to a few hundred effective founders (Carmi et al. 2014), and the medieval genomes from Erfurt show that the founder event and the modern ancestry profile were already in place by the fourteenth century (Waldman et al. 2022). Several other communities have their own, smaller bottlenecks. This matters for testing, as the next section explains.

Why qpAdm rather than a percentage tool#

A consumer percentage and a qpAdm model answer different questions, and for a Jewish genome the gap between them is wide. Here is what each one measures, and what to ask before you pay for a model.

What a percentage tool measures#

A consumer ancestry breakdown compares your genome with reference panels of living people, grouped by self-reported origin, and reports how much of your genome is best matched by each group. For most of the world that is a reasonable proxy for geography. For Jewish customers the reference groups include one or more Jewish categories, and there the method does something that is technically correct and practically misleading.

Because of the medieval bottleneck and centuries of endogamy, Ashkenazi genomes resemble one another more closely than the members of almost any other population do. The panels recognise that tight cluster with very high confidence and treat it as a category in its own right, so a person with four Ashkenazi grandparents is reported as very close to 100% "Ashkenazi Jewish". That is a correct statement about resemblance to a reference group and a completely circular statement about origins. It cannot tell you that roughly half of that ancestry is Levantine, because the Levantine and southern European halves are both inside the reference before the counting starts; for the same reason it cannot separate Italian from Levantine, Iranian from Canaanite, or Iberian from Moroccan. The Sephardic, Mizrahi and Yemenite categories some panels offer work the same way, with smaller clusters.

Three consequences follow.

1. "0% Middle Eastern" is not a finding. It means the Middle Eastern ancestry was assigned to the Jewish category, not that it is absent.
2. Partial results are unstable. A customer with one Jewish grandparent often sees the Jewish share drift between updates, because the panels are re-drawn and the bottleneck signal is easy to over- or under-count.
3. The tool cannot fail. There is no version of the calculation that comes back and says the proposed explanation is inconsistent with the data. Every genome gets a breakdown that sums to 100%.

Free calculators built on Global25 coordinates share the third property: they return a distance to the nearest population averages and a mixture that always sums to 100%. They are useful for exploration, and we offer Ancestry Map, a Global25 analysis, ourselves; the comparison with qpAdm explains which question each method answers.

What a qpAdm model measures#

qpAdm is the method from the ADMIXTOOLS package, developed in David Reich's laboratory, that most published ancient-DNA admixture studies have used since 2015, including the Levantine and Ashkenazi studies cited above. It takes a target genome, in this case yours, and asks whether it can be written as a mixture of a chosen set of source populations, tested against a fixed set of distant outgroups. It returns a weight for each source, a standard error on each weight, and a single p-value for the whole model. The p-value can reject the model. That is the property that separates a test from an estimate, and it is set out fully in Understanding qpAdm and the reading guide for the three numbers.

Against a percentage, three differences are concrete for a Jewish genome.

• The sources are ancient wherever the record allows. Instead of a reference group of living Ashkenazi Jews, the sources are populations such as the Bronze Age Canaanites of Megiddo and Hazor, the Imperial Romans of central Italy, Aegean Greeks, the Neolithic farmers of the Zagros, Iron Age Iberians, North African Berbers and Germanic and Slavic groups, depending on the community. Two Classical Antiquity sources, the southern Levant and the Arabian Peninsula, are the exception: the record holds almost no ancient genomes from those regions in those centuries, so they stand on present-day Levantine and Bedouin genomes, and the description of each in the report says so. The Levantine and European halves of an Ashkenazi genome are therefore separate sources with separate weights, and each weight's standard error says how confidently the model separated them. Where two neighbouring sources are close, more than one can fill the same slot, so a name in the report is the population the model used, not proof that no other would fit.
• The model can fail. If an Ashkenazi genome is proposed as Canaanite plus Imperial Italian and that mixture is inconsistent with the data, the p-value says so. If a third source is proposed that the genome does not need, its weight comes back indistinguishable from zero, and the simpler model without it, one of the nested models our analyst checks before publishing, still fits.
• Every number can be checked. Where a percentage tool prints one number per label, a qpAdm model comes with a standard error, a Z-score and a confidence interval for each weight, a p-value and a complete outgroup list, and our report puts all of them on screen.

A model that fits is a description, not a history. It gives your ancestry's composition relative to the groups that were sampled; it is not proof that those groups met and mixed. Where related populations shaded into one another across a region, more than one model can fit, and each is one adequate description of your genome rather than the story of how its ancestry arrived.

Since August 2026 our source catalog carries a dedicated Canaanite (2000 - 1200 BC) source, built from the Israel_MLBA genomes of Megiddo, Hazor and their neighbours, precisely because modelling a Jewish or Levantine genome without a Bronze Age Levantine reference forces the model to split that ancestry between Iron Age and Roman-era proxies that already carry Aegean and Anatolian admixture. The full source directory lists every population, with a page for each.

A checklist before you buy, anywhere#

Whoever you buy from, these questions tell a model you can check apart from a percentage.

• Does the product show a p-value, a standard error and Z-score per source, and the complete outgroup list? If not, it is not a qpAdm result you can check.
• Are the sources ancient populations, or modern Jewish reference groups relabelled? Where a source is a present-day stand-in, does it say so?
• Is the full record on screen, down to the chi-square, the SNP counts and the number of ancient individuals behind each source, so that another analyst could check it?
• Does it tell you what your file's coverage supports before you pay?
• Does it claim any population, tribe or person as your ancestor? It should not.
• Does it claim to measure who is Jewish? It cannot, and it should say so.

What an Ancient Ancestry report looks like for a Jewish genome#

Every report models two eras. The Hunter-Gatherer and Neolithic Farmer era writes your genome in terms of the deep ancestral streams of western Eurasia: Anatolian and Iranian farmers, Natufian foragers, steppe herders, European hunter-gatherers. For Jewish genomes of every community the Anatolian farmer, Iranian farmer and Natufian sources carry most of the weight, with the steppe and hunter-gatherer sources marking the European admixture in Ashkenazi and Sephardic genomes and absent or near zero in Mizrahi and Yemenite ones.

The Classical Antiquity era is where the communities separate, because the sources are closer in time to the events that made them. A typical Ashkenazi model uses the Canaanite source beside Imperial Italy, sometimes with a small Germanic or Slavic share; a typical Iraqi Jewish model uses Canaanite beside an Iranian-related source; a Yemenite model tests Canaanite against the Arabian Peninsula source; a Sephardic model tests Canaanite against Iberian sources, and against North African ones when the declaration includes a North African community. Each community's post describes its candidates in detail.

The published model is found the same way for every community. For Classical Antiquity the candidate sources come from a pool curated for your declared origins before any model runs (the deep-time era has few enough populations that all of them are searched); our analyst tests models from that pool, automated searches included, and re-tests the survivors; and a person reviews the candidates, chooses one model per era and approves it. That is the model you receive.

For every era the report's Your model view shows that model's record, under For specialists: seven fit figures (the p-value, chi-square, degrees of freedom, f4 rank, minimum SNPs per f4, merged SNPs and jackknife blocks) under a verdict such as "Clears Ancestrify's bar"; a Sources table giving each source's panel label, sample count, share, weight, standard error, Z-score and 95% confidence interval; and the right (outgroup) populations with their sample counts. The Atlas view is the map, and every order also gets two rendered videos, a Globe film and a Populations film, each covering every era. The nested models and the rank test are what our analyst checked before publishing; they are not printed in the report. Every number in the record is explained in The model record explained.

The publish bar#

Every model we publish must clear the same numeric bar in every era: a p-value above 0.05, and for every source a Z-score above 3 in magnitude and a standard error below 0.10. That bar is our own convention, not a criterion the audit literature validated. One property is worth stating: with every weight positive, a Z-score above 3 places each weight more than three standard errors from both 0 and 1. Every tier, from €29.99 to €59.99, gets the same report, maps and videos, the same model record and the same bar.

What it cannot do#

Three limits are worth stating before you buy, because they are specific to Jewish genomes.

1. It cannot resolve the bottleneck away. After a founder event a population drifts a little away from every ancient source at once, in a direction no source supplies. A well-composed model absorbs this; a poorly composed one produces a source with an inflated weight, or a rejection with no visible cause. What sets the error rate of a search over models is mostly its pre-study odds, how plausible its candidates were before testing, not whether a script or a person ran it (Flegontova et al. 2025). That is why the candidate sources are chosen for your declared origins before any model runs, and why an automated search can propose candidate models but cannot publish one: a person reviews the candidates and approves every published model.
2. It cannot separate closely related sources with a low-coverage file. Canaanite versus Iron Age Phoenician, Imperial Italian versus Aegean, Canaanite versus Arabian: these pairs are close in genetic space. The standard error that tells you how well a model separated them is set above all by your file's coverage, and no search can add SNPs your file does not have. A low-coverage file can return a model whose p-value passes while one source's standard error is above 0.10, and that model is not published. Check your file's coverage for free in the file check before paying.
3. It says nothing about identity, descent from any named person, tribe or lineage, or religious status. A source population is a statistical proxy, not a family, and a passing model is one the data did not refute, not one they confirmed. A weight on the Canaanite source is a share of your ancestry measured relative to the sampled Bronze Age Levantine genomes and the other sources in the model. It does not mean any individual buried at Megiddo was your ancestor, and it is not evidence for or against anyone's Jewishness.

After the report#

Once a report is published, its sample can be attached to the Ancestry Workbench for free, and from the Your model view you can open the published model there. The Workbench runs on a prepaid balance from €9.99 that never expires: you run your own qpAdm models on your own merged sample, with the same panel of ancient genomes, including the medieval Erfurt Jewish genomes and every Levantine group in the AADR, and each completed run is paid from the balance. Expect rejections; they are the method working, and failed runs are never charged. The walkthrough is in Run your own qpAdm models.

Where to start#

• Read the post for your community, then the Canaanite source.
• Check your raw file's coverage in the free file check.
• Order Ancient Ancestry once the file check shows your file can support it.
• If you want distances and a PCA rather than a formal model, Ancestry Map compares your coordinates with the modern Jewish community averages and the ancient Levantine samples directly, and Ancient Matches scans your genome for shared segments with individual ancient people.

References#

• Agranat-Tamir, L. et al. (2020). The genomic history of the Bronze Age Southern Levant. Cell, 181(5), 1146 to 1157.
• Atzmon, G. et al. (2010). Abraham's children in the genome era: major Jewish diaspora populations comprise distinct genetic clusters with shared Middle Eastern ancestry. American Journal of Human Genetics, 86(6), 850 to 859.
• Behar, D. M. et al. (2010). The genome-wide structure of the Jewish people. Nature, 466, 238 to 242.
• Carmi, S. et al. (2014). Sequencing an Ashkenazi reference panel supports population-targeted personal genomics and illuminates Jewish and European origins. Nature Communications, 5, 4835.
• Flegontova, O. et al. (2025). Performance of qpAdm-based screens for genetic admixture on graph-shaped histories and stepping stone landscapes. Genetics, 230(1), iyaf047.
• Haber, M. et al. (2017). Continuity and admixture in the last five millennia of Levantine history from ancient Canaanite and present-day Lebanese genome sequences. American Journal of Human Genetics, 101(2), 274 to 282.
• Waldman, S. et al. (2022). Genome-wide data from medieval German Jews show that the Ashkenazi founder event pre-dated the 14th century. Cell, 185(25), 4703 to 4716.
• Xue, J. et al. (2017). The time and place of European admixture in Ashkenazi Jewish history. PLoS Genetics, 13(4), e1006644.