Why Does My DNA Test Show North African Ancestry When My Family Is Spanish or Latin American
A small North Africa share in a Spanish, Portuguese, Canarian, Hispanic or Latin American DNA result is common. Here is when it is real history, panel overlap or noise.
A small North African share in a DNA result for someone with Spanish, Portuguese, Canary Islander, Hispanic or Latin American family is common and usually unsurprising. It comes from one of three things, often a mix: real Iberian history, Mediterranean ancestry that Iberia and North Africa share, or statistical noise at low percentages. Analysing a raw file from 23andMe, AncestryDNA or MyHeritage against a reference panel gives the share and its range. The file itself holds only genotypes, and it cannot tell you who the ancestors were.
Iberia's history put it there
Iberia was not a closed population. Muslim rule began in 711 and the last kingdom, Granada, fell in 1492. Moriscos, the Muslims who converted, were expelled between 1609 and 1614, but many people stayed, and their descendants simply became part of the general population. The genetic picture matches this. Bycroft et al. 2019 (Nature Communications) analysed people across Spain and found North African-related ancestry across most of the country, notably higher in the west and south-west, and lowest or near-absent in the Basque area. The Canary Islands add a second route: the original islanders were of Amazigh (Berber) origin from North Africa, and some of that ancestry survives in people from the islands today.
If your family is Hispanic or Latin American
Latin America inherits all of it. A large share of early colonists came from southern and western Spain, and Canary Islanders settled Cuba, Venezuela, Puerto Rico and Louisiana in numbers. Andalusia and Extremadura were overrepresented, especially early, but Castile and the north sent many people too. So a North Africa share can cross the Atlantic in a family that has never thought of itself as anything but Spanish or Mexican or Cuban.
Portugal, which settled Brazil, has a comparable history: Muslim rule in the south lasted until the Algarve was taken in 1249, so a small North Africa share is plausible in Portuguese and Brazilian results too.
Most Hispanic results also carry Native American and West African shares, and that matters here. North African and West African reference groups are not close on the map of genetic variation; North Africans cluster nearer Europe and the Middle East. Small shares can still leak between the two lines, because many North African populations carry some sub-Saharan ancestry, so the two components can trade small shares in a mixed genome. If you are Latin American, do not read the North Africa line alone. Check the North Africa and West Africa shares together, with their ranges, and ask whether one might be absorbing part of the other.
Three different things produce the same line on your report
The first is recent-ish history: mixing within the last 500 to 1,300 years. The second is older and less dramatic. Iberians and North Africans both descend partly from the same early farming populations and have traded genes across the Strait of Gibraltar for thousands of years. A reference panel has to draw a line between them, and the line is blurry. The third is noise. The software fits your genotype as a mixture of reference groups, and when your best match is Iberian but the Iberian reference is imperfect, a few percent leaks into the nearest neighbours. Small shares can also trade between North African and West African components, since many North African populations carry some sub-Saharan ancestry.
How to judge whether your share is probably real
The single most useful number is the range printed beside the percentage. A share of 4% with a range of plus or minus 1 is a different statement from 4% with a range of plus or minus 4. The second is compatible with zero.
| Example | Share | Range | Probably real? |
|---|---|---|---|
| A | 12% | ±3 | Likely |
| B | 5% | ±2 | Possibly |
| C | 4% | ±4 | Cannot tell from zero |
Scroll the table sideways for the rest →
Illustrative examples, not real results.
Things that make a share more believable:
- It is larger than its own range, and not by a hair.
- It shows up again on a different panel or company, with similar size.
- Your closest reference populations also include North African or other southern Mediterranean groups, not just Iberian ones.
- You have a documented family link, such as a grandparent from the Canary Islands or from southern Spain.
Things that make it less believable:
- It sits below 2 or 3% with a range that touches zero.
- It appears in one tool and vanishes in another.
- It is the only unexpected region, with no pattern around it.
This is a rule of thumb, not a test. Nobody can put a hard line at 2%, and how reference panels handle Iberian samples is still an active argument.
Why a raw file cannot say Moorish, Amazigh or Guanche
A raw file from a genotyping chip holds roughly 600,000 to 700,000 spots in your genome, each a letter. Whole-genome files hold millions of positions. Either way, they contain no labels. Moorish was a medieval European word for Muslims of North Africa and Iberia, a mixed group by any measure. Amazigh is a language and identity, not a sharply defined genetic block.
Guanche is the name for the pre-Hispanic people of the Canary Islands. Your genome cannot tell you that you descend from them. It can only show that you resemble certain reference groups, modern or ancient.
What comes out is similarity to named panels. Any tool that prints Moorish or Guanche on your report is giving you a story built on top of that similarity.
A small North Africa share is a measurement with a margin of error, not a label for your ancestors.
A North Africa share does not tell you your ancestors' religion
People reasonably ask whether the share means Muslim or Sephardic Jewish forebears. Sometimes history says so, but DNA does not read religion. A Levant share beside it fits some stories and a Mediterranean-wide baseline fits others, and the file alone cannot separate them. Where there is a paper trail, the paper trail wins.
The length of ancestry tracts (local ancestry) is the real tool for dating mixing, since recent mixing leaves long tracts and old mixing leaves short ones. Some companies show these as chromosome painting. HealthOSX shows neither tract lengths nor chromosome painting, so that dating tool is not on offer here. Relative-matching services are a different thing: they show segments shared with relatives, and they do not date admixture.
Seeing it in context
The percentage matters less on its own than in context. The free reader at healthosx.com/dna fits your file against 157 present-day populations across 15 world regions. North Africa, the Levant, Arabia, Anatolia/Iran and the Caucasus are separate regions rather than one West Asian bucket, which most consumer tests sold in the US and Europe do not offer. Each share comes with its likely range, so you can apply the test above directly. It also ranks your closest populations among 298 present-day groups. That ranking shows which groups you resemble most, and you can see whether North African groups appear beside Spanish ones. Resemblance to a Jewish group in that list is not evidence of Sephardic descent, for the reasons above.
The reader also draws a PCA plot, with you as one dot among the reference groups, and lets you download it as a CSV. It shows directly whether you sit between the Iberian and North African clusters, or inside the Iberian one.

Choosing a file. The report is read on your own device.

Ancestry across 15 regions and 157 reference populations.
If you have a raw file, upload it to the free HealthOSX DNA reader and compare your North Africa, Levant and West Africa shares, and their ranges, then look at the closest-populations ranking and the PCA plot for where you fall. Ancestry runs on the server and only with your explicit consent. It sends only the markers it needs, not the whole file, and you can delete them at any time with 'Delete my DNA data'. The trait and health story is read in the browser first.
New to raw DNA files? Start with the full guide: 23andMe Raw Data Interpretation Free in Your Browser
Written by
HealthOSX Research