Power and Limitations of Inferring Genetic Ancestry

ABSTRACT

Background

The recent emergence of technologies that capture and analyse genetic variation patterns obtained from a person's DNA sample has led to numerous academic and commercial endeavours to infer individuals' ancestries. In theory, a person's genome contains a wealth of readily accessible information regarding their ancestors, despite only some of our ancestors contributing to the DNA we carry. This makes genetic tests an attractive alternative to the painstaking reconstruction of family trees or directly contacting long‐lost relations, particularly when, unless there are notable individuals in the tree, historical and genealogical records tend to diminish in frequency with each generation. However, while powerful, there are limits to what genetic data can unearth, as well as important assumptions underlying these analyses.

Methods

This review describes some of the early history and latest advances in techniques and data used to infer ancestry using genetics, highlighting both the power and limitations of current studies.

Conclusion

While genetics is a powerful means of exploring aspects of people's ancestry, a stronger focus on conveying uncertainty will allow both academics and non‐academics to avoid the ever‐present risks of over‐interpretation.

1. Background

The term ancestry is problematic for a number of reasons. For example, it can mean our genealogical ancestors, i.e. those individuals whose identities we can trace using historical and genealogical documents. Many people conflate the terms ancestry and ethnicity. In particular, ethnicity can be considered to be a more fluid term based on how an individual or individuals claim identity with a particular group based on a shared culture, language, religion or history and can therefore change within a person's lifetime. Genetic ancestry is something else again as, due to how our DNA is inherited, not all of our genealogical ancestors are represented in the DNA that we carry. Therefore, genetic ancestry refers only to the subset of paths through your genealogical pedigree that you have inherited genetic material (see Mathieson and Scally ²⁰²⁰ for a review of different definitions of ancestry).

In this review, we will confine ourselves to genetic ancestry. We discuss recent advances in studies of genetic ancestry, while also highlighting some of the limitations and assumptions of these studies. These limitations include the widely‐used practice of categorising individuals into distinct groups, e.g. using genetics or ethnic labels, which—rather than having any biological relevance—is done for convenience to assist analyses and interpretations. The potential for such classifications to propagate harmful ideas regarding “biological race”, along with the increasing visibility of genetic data to the general public through e.g. DTC tests, underscores how effective communication of the field's limitations, in addition to its potential, are essential.

2. Data Types and Initial Studies

Due to the simplicity of their inheritance patterns, initial ancestry analyses focused on studying mitochondrial DNA (mtDNA), which is passed from mother to offspring, and the non‐recombining portion of the Y chromosome (NRY), which is passed from father to son. In 1987, Cann et al., using low‐resolution RFLP analysis, examined the mtDNA types of 147 individuals, and from this, inferred that the common ancestor of the research participants was a woman who lived probably in Africa some 200,000 years ago (Cann et al. 1987). On publication, it generated international headlines and within months, Cann had dubbed this individual ‘Mitochondrial Eve’: she was known as the ‘mother’ of all living humans.

As mtDNA analysis developed, focusing first on sequencing the two hypervariable regions HVS1 and HVS2, further mtDNA types, known as haplogroups, were determined. By 1996, geneticist Bryan Sykes had set up a commercial DNA testing company ‘Oxford Ancestors’, offering the paying public an analysis of their mtDNA type. Further interest in his company, and mtDNA testing in terms of genetic ancestry, was boosted by the publication of his book ‘The Seven Daughters of Eve’ which, riffing on that early paper, artificially split mtDNA types into seven groups, which he gave the names of fictional regional ‘daughters’ (Richards et al. 1996; Sykes ²⁰⁰¹). Here the use of genealogical terms and names fuelled the public imagination and was the start of the direct‐to‐consumer (DTC) genetic ancestry testing industry we see today.

Testing of mtDNA types for genetic ancestry was followed swiftly by a similar analysis of Y chromosomes. Here, again, Bryan Sykes was one of the early providers of such testing for the public, first using a set of 10 short‐tandem‐repeat (STR) markers, short repetitive stretches of DNA sequence, to predict a Y chromosome haplogroup—he also gave ‘clan’ names and determined the testee's ‘ancient ancestral father’. Here the terminology is used to conjure up a closer relationship that exists—using the term ‘father’ for someone who would have lived many generations ago, some millennia in the past.

Analyses of mtDNA and NRY data thus heralded the start of the field of genetic genealogy. Of course, due to the inheritance pattern of these two segments of DNA, women could only take an mtDNA test, whereas men can take both a Y chromosome and a mitochondrial test. Another company, Family Tree DNA, soon offered similar tests, progressing over time to larger sets of STRs on the Y chromosome to type. These could then be used to test within groups of men suspected to share a common ancestor, for example, due to sharing the same surname, to determine groups descended from a common ancestor within a genealogical timeframe or to prove or disprove family relationships.

While some mtDNA and NRY analyses have proven useful scientifically (King et al. 2006; King et al. ²⁰⁰⁷; King and Jobling ²⁰⁰⁹; King et al. ²⁰¹⁴), despite some of the over‐interpretations mentioned above, recent ancestry studies instead analyse the autosomal chromosomes, which comprise the vast majority of our genome. Key to studying genetic ancestry using autosomes is the process of recombination. Effectively, when an autosomal chromosome is passed from a parent to their offspring, each recombination event will bifurcate the chromosome into segments inherited from a different grandparent on that parent's side. Stacking these recombinations up over generations, a person's autosomal genome can be depicted as a mosaic of segments inherited from different ancestors who lived at some point in the past. Therefore, an important consequence of recombination is that a person's autosomes contain information from each of the many ancestors that passed genetic material through the generations to that person. In contrast, each of the NRY and mtDNA contain information from only one of these ancestors and hence represent a very incomplete picture of their progenitors, despite some early out‐sized claims of their importance (Figure 1).

FIGURE 1.

Mitochondrial DNA (blue fill) and Y chromosomes (orange outline) are each inherited from a single ancestor matrilineally or patrilineally, respectively. They contain no genetic information from any other ancestors (grey figures with black outlines). The autosomes are represented by patterned rectangles, with one autosomal chromosome pair depicted per person. One chromosome in a pair is inherited from each parent, and the process of recombination affects the amount of DNA inherited from each ancestor at each generation. Relative to the mtDNA and NRY, the autosomal chromosome pair contains genetic material from many of an individual's ancestors, hence providing a more complete picture of their genetic history.

Even across the autosomes, many of the people in one's distant family tree are expected to contribute little or no DNA to that person. For example, while each of a person's four grandparents is expected to contribute to 1/4th of their autosomal genome, ten generations (~250–300 years) ago, a person can have at most 2^10 = 1024 ancestors. Thus each of these more distant ancestors is expected to contribute to <1/1000th of their autosomal genome, with a ∼58% chance of no detectable relationship at all. At fourteen generations ago (∼420 years), this probability goes up to >95% (Donnelly 1983; Ralph and Coop ²⁰¹³). For this reason, our genome is necessarily only a snapshot of a fraction of a person's complete set of ancestors and carries no information about the ancestors that did not contribute any DNA to this particular descendant.

3. Inferring Shared Recent Ancestry Using Autosomal Data

Due to recombination, individuals who are more closely related are expected to share longer segments of autosomal DNA with each other. For example, grandparents are expected to share longer segments with their grandchildren than great‐grandparents are with their great‐grandchildren. Similarly, cousins are expected to share longer segments than second cousins, as cousins' shared segments are inherited through a common grandparent and second cousins' shared segments are inherited through a common great‐grandparent. Given the mutation rate per meiosis in humans is small, such segments inherited from a shared recent ancestor are expected to contain the same (or very similar) genetic sequence data, with most studies today analysing biallelic single‐nucleotide polymorphism (SNP) positions where a given segment can carry one of two possible alleles (genetic variants). Based on this logic, several programs e.g. hap‐IBD, ancIBD and ibd‐cluster (Browning and Browning 2024; Ringbauer et al. ²⁰²⁴; Zhou et al. ²⁰²⁰) try to identify long segments (e.g. >2Mb) for which two individuals share matching SNP data, as these may indicate so‐called identical‐by‐descent (IBD) segments that both have inherited from the same recent ancestor. The total amount and lengths of IBD sharing genome‐wide can then be used to infer the two individuals' degree of kinship (Seidman et al. 2020). Indeed some DNA testing companies return results showing the inferred amount (measured in centimorgans) of genome‐wide DNA shared between a customer and others in the reference database (Guerrini et al. 2022). Using currently available genetic variation datasets in modern humans, IBD‐inference programs typically can reliably estimate kinship up to a 6th degree (Ramstetter et al. 2017).

Identifying close relatives in this manner is of enormous interest in many applications, such as DTC services, forensics and reconstructing pedigrees using genetic data from archaeological sites (Cassidy et al. 2025; Tillmar and Kling ²⁰²⁵). However, most academic and commercial studies of ancestry instead use collections of unrelated people, where sampled individuals have little inferred IBD (i.e. long segment) sharing. This is because unrelated people share more distant ancestry that is often not well‐captured by other non‐genetic records. To learn about a person's ancestry, several programs such as Chromopainter, RFmix, MOSAIC and supervised ADMIXTURE (Alexander and Lange 2011; Lawson et al. ²⁰¹²; Maples et al. ²⁰¹³; Salter‐Townshend and Myers ²⁰¹⁹) compare the genetic data from a target individual of unknown ancestry to that in a set of reference individuals for which ancestry is presumed known. The aim of such analyses is to identify the person within the reference set that a target individual is most closely related to in a genomic region, i.e., this represents the reference person for which the target individual shares a more recent ancestor than they do with any other person in the reference set. Due to historical recombination as described above, this “closest relative” changes multiple times along each autosomal chromosome of a target individual (Figure 1). Typically, and intuitively, the closest relative in a genetic region is inferred to be the reference individual who carries the longest stretch of SNP data matching that of the target individual in the genetic region (Lawson et al. 2012). While this is a powerful approach to inferring genetic ancestry, other commonly used approaches perform matching using summaries of SNP data, e.g., projecting people with unknown ancestry onto principal components extracted from reference sample data (Herrando‐Pérez et al. ²⁰²¹; Patterson et al. ²⁰⁰⁶; Price et al. ²⁰⁰⁶).

4. Limitations of Autosomal‐Based Ancestry Inference

In addition to the methodology used for inference, this approach to determining a person's ancestry depends greatly on the reference set of individuals that the target is compared to. Typically the reference set consists of people sampled from particular geographic regions and/or people who have provided ethnicity or birthplace information. Reference individuals are then classified into different groups based on this geographic and/or ethnicity information. Sometimes genetic clustering algorithms are employed in an attempt to ensure these reference groups are relatively genetically homogeneous. A target individual that matches someone in a particular reference group is assumed to have at least some ancestors who were closely related to people from that reference group. While this makes logical sense, interpreting what this close relatedness means is not straightforward. Even the best‐matching reference individuals may be genetically distant from their true ancestors, which can occur if no good representatives for a person's ancestors exist in the reference set. As an example, many available reference datasets contain a larger number of people of European descent relative to those of non‐European descent (Adebamowo et al. 2022; Fatumo et al. ²⁰²²). In such cases, more precise matches, i.e., matching to reference individuals in fine‐scale geographic regions, are expected when inferring the ancestry of someone of European descent relative to someone with few recent European ancestors.

Furthermore, matching a person's genetic variation patterns to those of people in a particular reference group does not mean that the person's ancestors were ever a member of that current ethnic group or ever lived in that geographic region. This is an assumption, as genetics contains no geocoding or ethnicity information, and the genetic ancestry of people in a certain region can dramatically change through time with migration and drift. Conversely, a target individual may not match a reference group even if they have ancestors who are from the ethnic group or geographic region represented by that reference group. This can occur if the reference individuals representing that ethnic group or geographic region do not well reflect the complete genetic diversity of that group or region. For example, a reference set of British individuals that consists only of people sampled from a small geographic area within Wales may not be matched by someone else with recent British ancestry. Instead, this person may share more best matches with reference individuals sampled from France who happen to have a greater amount of genetic diversity, resulting in misleading inference. In addition to these issues, sample size bias can occur, whereby geographic regions and/or ethnic groups represented by more reference individuals are a priori more likely to be matched unless careful steps are taken to avoid this (e.g. ensuring all reference groups have the same number of people).

5. Recent Advances in Acquiring and Analysing Genetic Variation Data

Notwithstanding these complications, both the genetic data acquired and the methods used to analyse these data have made continuous strides. Of particular note, large‐scale whole‐genome‐sequencing (WGS) datasets, which consist of high‐quality variant calls at all (millions of) SNPs within thousands to hundreds of thousands of individuals, are becoming increasingly available (All of Us Research Program Genomics Investigators ²⁰²⁴; Bycroft et al. ²⁰¹⁸; Milani et al. ²⁰²⁴; Nagai et al. ²⁰¹⁷; Walters et al. ²⁰²³). Such datasets enable finding which individuals share variants that are very rarely found in the rest of the population, such as doubletons (a genetic variant found in only two individuals within a sample), which are indicative of very recent ancestor sharing (Mathieson and McVean ²⁰¹⁴). Emerging technologies are increasingly accurate at interrogating parts of the genome that were previously difficult or impossible to capture, enabling telomere‐to‐telomere assembly that includes better assessment of (e.g.) copy‐number‐variants and ribosomal DNA patterns that may further improve ancestry inference (Aganezov et al. 2022; Miga and Eichler ²⁰²³; Nurk et al. ²⁰²²).

In addition, the technology to reliably sequence DNA from ancient human remains (aDNA) is continually improving (Liu et al. 2022). Consequently, recent years have seen a massive increase in the amount of available aDNA representing worldwide geographic regions over different time periods (Akbari et al. 2024; Mallick et al. ²⁰²⁴). In many cases, relative to present‐day samples, aDNA can offer a better genetic representative of the ancestral groups that a target individual descends from. When using present‐day reference samples to learn whether a target individual has ancestors from genetically different (e.g. geographically distant) groups, a major challenge is that many people alive today themselves descend from the intermixing of genetically different groups. DNA samples from individuals that predate these intermixing events can circumvent this issue. For example, ancient genomes from Jomon hunter‐gatherers have been useful for understanding patterns of ancestry in modern Japanese people who are admixed between this ancient population and continental East Asian groups. Similarly, genomes from Viking age individuals have been used to investigate their genetic legacy in modern‐day Europeans, and present‐day Ethiopians’ genomes have been shown to have differing amounts of ancestry related to a 4500‐year‐old ancient Ethiopian forager (Watanabe and Ohashi 2023; Yamamoto et al. ²⁰²⁴; Margaryan et al. ²⁰²⁰; López et al. ²⁰²¹). Of course, the complexity of human history has demonstrated that aDNA samples often descend from intermixed groups too, and challenges in acquiring reliable data from particular geographic regions and time periods can still hinder sampling the “best” ancestral proxies.

Historical samples (from within the last millennia) are increasingly being published, including those from famous individuals, such as Beethoven, King Richard III of England and Lakota Sioux leader Sitting Bull (King et al. 2014; Moltke et al. ²⁰²¹; Begg et al. ²⁰²³). Some DTC companies, such as 23andMe, DNA Consultants and MyTrueAncestry now allow individuals to compare their DNA to certain historical individuals and ancient DNA samples.  However, such mtDNA or Y chromosome matches are likely to be on timescales that are not genealogically relevant. Moreover, given the doubling of ancestors in each generation backwards in time, the importance of a genetic match to any specific historical individual on the autosomes may carry little significance (Chang 1999; Derrida et al. ²⁰⁰⁰; Ralph and Coop ²⁰¹³). Indeed any autosomal DNA sharing between a customer and a historical individual detected by these companies will almost certainly be small and due to them sharing some distant common genetic ancestor(s), rather than the customer directly descending from that person. https://www.forbes.com/sites/jenniferraff/2019/04/09/genetic-astrology-when-ancient-dna-meets-ancestry-testing/#6dcb5b496c69

High‐quality WGS data have also stimulated the advance of approaches that infer full (genetic) genealogies relating thousands of people back in time through their shared ancestors e.g., Relate, tsinfer, argWEAVER (Hubisz et al. 2019; Kelleher et al. ²⁰¹⁹; Speidel et al. ²⁰¹⁹; Wong et al. ²⁰²⁴). These approaches leverage the full set of variants that samples carry across the genome as a clock to infer the number of generations ago each ancestor lived. In principle, the complete genealogy of sampled genomes represents the best obtainable information from genetic variation data. Computational considerations currently require a number of simplifying assumptions that make inferred genealogies deviate from the true ones to an unknown degree, while also limiting the number of samples that can be analysed jointly. However, these techniques are continually improving, and currently show unprecedented power to resolve aspects of demographic and evolutionary history (Lewanski et al. 2024; Wong et al. ²⁰²⁴; Speidel et al. ²⁰¹⁹; Lewanski et al. ²⁰²⁴; Wong et al. ²⁰²⁴).

6. Inferring Genetic Structure and Demographic History Using DNA

Along with these continuing advances, recent genetic analyses have provided new insights into human demography, including genetic links among people. Such research has demonstrated a notable correlation between genetic variation patterns and geography, both at continent‐level scales as expected (Rosenberg et al. 2002) but also within many countries (Byrne et al. 2020; Elliott et al. ²⁰²²; López et al. ²⁰²¹). Some of this genetic structure is remarkably fine‐scaled. For example, when incorporating sufficient reference samples, genetic variation data from arrays with hundreds of thousands of SNPs can fairly reliably distinguish people from different counties within England (Hu et al. 2025) and from different ethnic groups living <20 km apart within Cameroon (N. Bird et al. ²⁰²³). This likely reflects how people historically on the whole tend to intermix with people that live near them, making it possible that such clusters of individuals share several recent ancestors who lived in or near that area.

As noted above, new techniques to infer genetic genealogies using WGS data have enabled more precise inference of the time when two people first share a common ancestor at locations throughout the genome (DeHaas et al. ²⁰²⁴; Kelleher et al. ²⁰¹⁹; Speidel et al. ²⁰¹⁹; Wong et al. ²⁰²⁴). This in turn enables better inference when populations split from one another. Such approaches have been used to infer the date of the initial ‘out‐of‐Africa’ dispersal of anatomically modern humans that have left present‐day descendants, as well as splits between non‐Africans in Europe and Asia and between different groups within Africa (Bergström et al. 2020). aDNA can also be used to infer splits between ancient groups, such as between different hunter‐gatherer groups in West Eurasia (Speidel et al. 2021). It is important to note that split times tend to represent an ‘average’ (midpoint) split for the lineages and splits between populations are often more gradual, perhaps over many thousands of years. In addition, there is increasing evidence that in some cases assuming a tree‐like model of splits oversimplifies human history, and a more structured model with multiple populations changing connectedness through time may be closer to the truth (Ragsdale et al. 2023).

These approaches can also be used to estimate changes in the effective population size of a group over time (Li and Durbin 2011; Schiffels and Durbin ²⁰¹⁴; Steinrücken et al. ²⁰¹⁹; Terhorst et al. ²⁰¹⁷). For example, time periods where lots of people from a sampled group are inferred to share common ancestors can indicate population bottlenecks that occurred during those times. Conversely, time periods where there are few shared ancestors suggest population expansions or the formation of population structure (Hilgers et al. 2025). The humans who first migrated out of Africa experienced a significant bottleneck in population size which can be still detected today using such concepts. In contrast, within the past 10,000 years, the majority of human populations have experienced major growth, although some African hunter‐gatherer groups appear to be exceptions (Bergström et al. 2020; Fan et al. ²⁰²³).

Other approaches that infer more recent (i.e. within the last 3000 years) changes in effective population size are also available such as IBDNe and GONE (Browning et al. 2018; Browning and Browning ²⁰¹⁵; Novo et al. ²⁰²³; Tournebize et al. ²⁰²²). Recent bottlenecks have been inferred in many present‐day populations, for example, Andamanese Islanders, Ashkenazi Jews, Finnish and Hispanic populations. New methods also infer effective population size changes in ancient individuals, with papers reporting a population expansion in Vikings during the Iron Age and in Medieval Britain before the Black Death (Fournier et al. 2023; Huang et al. ²⁰²⁵).

In addition to these insights, the process of recombination can be leveraged to learn about if and when populations intermixed (i.e., admixed) in the past. In particular, when two or more groups intermix, their descendants’ autosomes will be a mixture of segments inherited from ancestors representing each admixing group (Falush et al. 2003). Since the rate of recombination per each meiosis event is well characterised in humans (Auton et al. 2015; Palsson et al. ²⁰²⁵), the size of contiguous segments inherited from each ancestral group carries information on precisely how many generations ago the groups intermixed. This idea is leveraged by several available programs, such as ROLLOFF, ALDER, fastGLOBETROTTER, MOSAIC and DATES (Chintalapati et al. 2022; Hellenthal et al. ²⁰¹⁴; Loh et al. ²⁰¹³; Moorjani et al. ²⁰¹¹; Salter‐Townshend and Myers ²⁰¹⁹; Wangkumhang et al. ²⁰²²).

By dating admixture and combining this with historical or archaeological sources, we can learn more about events in human history, such as migrations. In Europe, the spread of agricultural practices was accompanied by admixture between migrating farmers and local hunter‐gatherers, and the inferred dates of these mixing events reveal the complex nature of the expansion, with farmers migrating along multiple routes (Chintalapati et al. 2022). A similar, complex, picture is found when inferring admixture dates between Bantu‐speaking agriculturalists and local hunter‐gatherers in sub‐Saharan Africa (N. Bird et al. ²⁰²³; Fortes‐Lima et al. ²⁰²⁴; Tallman et al. ²⁰²³). Admixture events are also often inferred to overlap with periods of historical trade or large empires/states, which likely precipitate the meeting of people from different geographical regions (Brielle et al. 2023; Hellenthal et al. ²⁰¹⁴; Schlebusch and Jakobsson ²⁰¹⁸; Wangkumhang et al. ²⁰²²). Incorporating Y chromosome and mtDNA data can also help determine whether some of these migrations were sex‐biased (Bajić et al. 2018; Wood et al. ²⁰⁰⁵).

7. Ethical Considerations

There are several important ethical considerations around the use of genetic variation data. First, it is inherently identifiable and can contain information about sensitive details, such as an individual's physical traits, susceptibility to disease and genetic ancestry (Fox 2020; Joly and Dalpe ²⁰²²; Mohammed Yakubu and Chen ²⁰²⁰; Thomas et al. ²⁰²⁴). Such data therefore can pose security and privacy concerns without suitable protective measures in place. Importantly, an individual's genetic data has identifiable information about their family too. Publicly available genetic databases, such as GEDmatch, allow the identification of specific individuals using long‐range familial relationships. One study projected that around 60% of U.S. individuals of European descent could be re‐identified through distant family DNA in a public database (Erlich et al. 2018). This technique has been successfully used by law enforcement to identify suspects in criminal cases, but also has the potential for misuse (Greytak et al. 2019; Kennett ²⁰¹⁹; Kuru ²⁰²⁵). In addition, there are concerns over the collection of genetic data from people without adequate consent, and the unethical use of genetic databases for surveillance (D'Amato et al. ²⁰²⁴; Lipphardt et al. ²⁰²¹; Moreau ²⁰¹⁹; Normile ²⁰²¹; Wessel ²⁰¹⁹; Whitmore et al. ²⁰²³).

As with studies linking genetic variants to traits such as disease, genomes with diverse representations from different worldwide groups are required for research into genetic ancestry, especially in regions that have been historically underrepresented in genomic research (Fatumo et al. 2022). Recently, there have been increased calls for a more diverse representation of populations in genomic databases (Martin et al. 2017; Popejoy and Fullerton ²⁰¹⁶). However, resource, capacity and specialist knowledge inequalities between the Global North and other regions of the world mean that such genomes are often sequenced and analysed in the former, potentially leading to imbalanced relationships (Ali et al. 2021; Atutornu et al. ²⁰²²). Related to this, criticisms regarding ‘parachute’ research and allegations of research malpractice have fostered a distrust of genetic research in several communities (Claw et al. 2018; Nordling ^{2018, 2017}; TallBear ²⁰¹³). An ideal solution is the creation of equal and meaningful partnerships between collaborators, where all contributions to a project are considered equally valid. Guidelines also emphasise the importance of community engagement, effectively communicating research findings and capacity building and training endeavours (Adame 2021; Haelewaters et al. ²⁰²¹; Martin et al. ²⁰²²; Sawchuk et al. ²⁰²⁴).

The rapid increase in sequencing genetic material from ancient samples brings its own ethical issues (Alpaslan‐Roodenberg et al. ²⁰²¹; Kowal et al. ²⁰²³). One issue centres on consent; while the deceased cannot provide consent, there may be descendants or culturally affiliated communities who can be consulted before the research begins. Such community consultation and engagement is also important to discuss the potential negative consequences aDNA ancestry research could bring (Bardill et al. 2018; Cortez et al. ²⁰²¹; Gibbon ²⁰²⁰; Sawchuk et al. ²⁰²⁴). Additional ethical issues arise from the fact that human remains must be damaged to extract DNA from them (Prendergast and Sawchuk 2018), and the possibility of identifying living descendants from ancient and historical DNA samples (Harney et al. 2023).

Finally, an ongoing ethical concern with research into genetic ancestry is its potential for reinforcing and propagating outdated ideas of ethnic and racial ‘essentialism’ (K. A. Bird and Carlson ²⁰²⁴). The use of genetic ancestry categories, widely used by researchers for convenience, can be misinterpreted by non‐experts to reflect discrete, genetically homogenous ‘pure’ sources, ignoring the complex demographic histories and intermixing discussed in this review (Lewis et al. 2022; Saini ²⁰¹⁹). The criteria used to select some reference samples may add to this problem. For example, one of the earliest collections of DNA from worldwide populations, the Human Genome Diversity Panel (HGDP), which is used in many genetic ancestry studies today, focused specifically on collecting DNA from ‘populations of anthropological interest’, i.e., those that were assumed not to have experienced significant migrations or admixture since the 16th century (Cavalli‐Sforza ²⁰⁰⁵). Such subtleties, and the limitations of methods for understanding genetic ancestry discussed above, can be difficult to communicate, or even be deliberately misinterpreted, with the increased availability of DTC tests seeming to exacerbate these issues (Carlson and Harris 2020; Kampourakis and Peterson ²⁰²³; Panofsky et al. ²⁰²⁰). Some research suggests scientific racism and eugenics‐related ideas have reached more mainstream political discourse in recent years, and scientists are becoming increasingly aware of how their work can be misappropriated (Wojcik 2025; Carlson et al. ²⁰²²). There have been calls for scientists to represent the complex truth of genetic ancestry more accurately, for example, by emphasising the continuous nature of genetic variation (e.g. with sampling strategies and plots), avoiding arbitrary ancestry categories and communicating limitations clearly (Birney et al. 2021; Committee on the Use of Race, Ethnicity, and Ancestry as Population Descriptors in Genomics Research et al. ²⁰²³; Coop ²⁰²²; Lewis et al. ²⁰²³).

8. Conclusion and Future Perspectives

The field of genetic ancestry studies has developed significantly since the early days of analysing mtDNA and Y chromosome diversity. New techniques that leverage the rich information in autosomal genetic variation data are enabling inference of distant kinship relationships and complex demographic processes such as population splits, changes in population size and admixture events. This coupled with the continuing increase in available data, both from modern and ancient individuals and from historically underrepresented groups, is allowing us to examine genetic history in even greater detail. As data and computational techniques in this field are still proliferating rapidly, the near future should bring increasingly more precise insights in this area. However, limitations still remain. Genetic ancestry information is complex, typically difficult to capture and often even more challenging to communicate to the general public. As with other fields, perhaps a stronger focus on conveying uncertainty will allow both academics and non‐academics to avoid the ever‐present risks of over‐interpretation. This in turn may lend more credence to reports highlighting the interesting facets of ancestry that genetics can reveal with high confidence.

Author Contributions

N.B., T.K. and G.H. wrote the manuscript. N.B. generated the figure.

Data Availability Statement

The authors have nothing to report.