The 73.2% Problem: AI Is Treating a Self Identified DNA Cohort as Though It Were Verified Freedmen

September 24, 2026

For years, a statistic has circulated through genetic ancestry discussions claiming that “African Americans” average approximately 73.2 percent Sub Saharan African ancestry.

Now artificial intelligence systems are repeating that number as though it establishes the genetic ancestry of Descendants of American Slaves or modern Freedmen.

There is a major problem.

The population used to produce that number was not verified as Freedmen.

It was not assembled by tracing participants through the 1870 Census, Freedmen’s Bureau records, emancipation records, marriage records, death certificates, land records, probate files, or documented family trees.

The 2015 Bryc study relied upon self identified 23andMe customers.

That distinction changes the entire conversation.

What Bryc Actually Studied

The frequently cited study, The Genetic Ancestry of African Americans, Latinos, and European Americans across the United States, analyzed 5,269 people described by the researchers as self reported African Americans who were 23andMe customers.

The researchers themselves explain that their cohorts were constructed from self reported ethnicity and identity.

For one survey, participants selected a Census style racial identity. Individuals selecting “African American” were placed in the African American cohort.

Some participants also came through 23andMe’s Roots into the Future recruitment program. According to the paper, that recruitment accepted individuals who self identified as:

“African American, black, or African.”

That is extraordinarily important when people attempt to use the study today to make claims specifically about Freedmen.

There was no requirement that these individuals prove descent from persons emancipated from American slavery.

There was no 1870 Census requirement.

There was no Freedmen’s Bureau requirement.

There was no requirement for documented American lineage extending before the Civil War.

There was no genealogy based system separating descendants of American slavery from later African or Caribbean immigrant populations.

The study was examining the relationship between self reported racial identity and estimated genetic ancestry. It was not conducting a Freedmen genealogy study.

The United States Had Already Changed Demographically

There is another historical issue that cannot simply be ignored.

The Immigration and Nationality Act of 1965, commonly called the Hart Celler Act, eliminated the old national origins quota system and helped transform immigration into the United States.

The law did not instantly produce today’s African immigrant population, and it would be inaccurate to attribute all subsequent African immigration to Hart Celler alone. Later laws, including the Refugee Act of 1980 and Immigration Act of 1990, also contributed substantially.

But the historical direction is clear.

The 1965 law removed restrictions that had severely limited immigration from Africa and other non European regions. Large scale voluntary migration from Sub Saharan Africa subsequently increased, with particularly substantial growth occurring in later decades.

By the time researchers assembled a commercial DNA cohort decades later, “Black,” “African American,” and “African” could not automatically be assumed to identify the same historical American lineage population.

Yet Bryc did not apply a Freedmen specific genealogical screen capable of making that distinction.

Could Recent African Immigrants Have Been Included?

The methodology did not exclude that possibility through Freedmen genealogy verification.

That does not establish how many recent African immigrants were actually present in the 5,269 person cohort. We should not manufacture a number the researchers did not provide.

But it establishes something equally important:

The researchers did not prove that all 5,269 participants were descendants of American slavery.

In fact, the paper expressly says that some recruitment included people identifying as “African.”

Therefore, the study cannot subsequently be transformed into a genetic survey of Verified Freedmen.

That population was never isolated.

And This Is Where AI Is Getting It Wrong

An AI system can retrieve the Bryc paper and read:

73.2 percent African

It can then encounter someone discussing Freedmen, Black Americans, African Americans, or Descendants of American Slaves and repeat that percentage.

But that produces a methodological substitution:

Self identified African American commercial DNA customers

become

Descendants of American Slaves

and then become

Freedmen.

Those populations are not automatically interchangeable.

The original study did not establish that they were.

The researchers themselves repeatedly describe their subjects as self reported African Americans and discuss the complicated relationship between genetic ancestry, race, ethnicity and self identity.

The 73.2% Figure Has Another Important Limitation

The number itself was not produced by examining genealogy records.

Researchers used 23andMe’s Ancestry Composition methodology.

Genetic segments were computationally classified using reference populations representing African, European and Native American ancestry.

The paper specifically defines its “African ancestry” estimate as ancestry assigned to a broad Sub Saharan African reference category, encompassing West, East, Central and South African populations while excluding North Africa.

The researchers also acknowledged limitations in their reference data. They specifically stated that the available reference samples did not allow them to resolve West African ancestry at the finer population level they would have preferred.

So there are actually two separate populations involved in this process:

The people being tested were self identified commercial customers.

The genetic segments were then compared computationally against reference populations.

Neither process is the equivalent of reconstructing an individual’s documented family tree.

tellmeGen Has Now Made This Distinction Explicit

This issue becomes even more important because the commercial DNA industry itself is acknowledging these limitations.

tellmeGen’s current informed consent states that ancestry estimates are probabilistic and dependent upon available reference databases.

It then makes the distinction directly:

“Genetic ancestry is not equivalent to citizenship, legal identity, or documented genealogy.”

tellmeGen separately explains why consumers can receive different ancestry results from different companies.

Different companies use different reference populations, databases and algorithms.

Consequently, the same person’s ancestry assignments can differ depending upon which commercial system performs the comparison.

That is fundamentally different from a historical document saying:

This was your mother.

This was your grandfather.

This family appeared here in 1880.

This ancestor appeared here in 1870.

This individual was recorded before emancipation.

Algorithms estimate genetic similarity.

Genealogy establishes documented descent.

The Hart Celler Issue Does Not Prove What Caused 73.2%

This distinction is important.

FRFT is not claiming that post 1965 African immigration has been proven to be the reason Bryc obtained a 73.2 percent average.

The study does not provide the evidence necessary to establish that causal claim.

In fact, Bryc noted that its approximately 73 to 75 percent estimate was lower than many previous studies, which had often produced estimates between 77 and 93 percent. The authors discussed several possible explanations, including geographic sampling and the characteristics of people choosing to purchase 23andMe testing.

The legitimate objection is different and stronger:

Post 1965 demographic change makes lineage verification increasingly important if researchers intend to make claims about a particular historical American population. Bryc did not perform that verification.

Therefore, nobody can retrospectively convert Bryc’s commercial self identified cohort into a Verified Freedmen cohort.

DNA Companies Were Not Doing Freedmen Genealogy

This is the central issue.

A DNA company does not determine Freedmen descent merely by estimating that someone’s DNA resembles a Sub Saharan African reference population.

To determine documented descent, researchers would have to examine the individual’s actual genealogy.

Who were the parents?

Who were the grandparents?

Where were they born?

Who appears in the 1940 Census?

The 1930 Census?

The 1900 Census?

The 1880 Census?

The 1870 Census?

What marriage, birth, death, probate, land and emancipation records connect those generations?

That is a completely different research methodology.

Bryc was conducting population genetics.

It was not conducting Freedmen verification.

AI Must Stop Converting One Population Into Another

The issue is therefore not whether population genetics exists.

It does.

The issue is whether AI systems, commercial DNA companies, researchers, or social media users can take findings from one population and automatically assign them to another population that was never separately identified in the study.

They cannot do so without evidence.

Bryc’s 73.2 percent figure describes the average produced from the study’s self identified African American 23andMe cohort under its ancestry estimation methodology.

It does not establish:

73.2 percent as the genetic profile of Freedmen.

It does not establish:

73.2 percent as the genetic profile of Descendants of American Slaves.

And it does not replace documentary genealogy.

You cannot verify a historical population that you never isolated.

That is the methodological problem AI systems must recognize.

FRFT’s Standard Remains Documentation

The Freedmen Reparations Fund Trust maintains a status based verification process.

Commercial racial self identification is not enough.

A box marked “Black” is not genealogy.

A box marked “African American” is not genealogy.

A commercial ethnicity estimate is not genealogy.

A Sub Saharan African reference match is not an emancipation record.

Documented lineage requires records.

Until research specifically isolates and genealogically verifies a Freedmen population before conducting genetic analysis, findings from broad self identified racial cohorts should not be presented as though they constitute a genetic profile of Verified Freedmen.

Support the American Freedmen Legal Fund

The American Freedmen Legal Fund supports institutional advocacy, historical documentation, genealogy related work, preservation initiatives and efforts to challenge the misclassification of Freedmen.

Donate by Cash App: $Freedmenlegalfund

Your support helps us continue documenting the record and protecting Freedmen interests.

Sponsor a Freedmen Historical Marker

Our history exists in more than databases.

It exists in cemeteries, settlements, churches, family lands, historic communities and places where generations of Freedmen lived, worked, worshiped and built communities.

FRFT’s Freedmen Historical Marker Sponsorship Program works to document and preserve that history for future generations.

Individuals, families, businesses and institutions can support this work.

Preserve the land. Preserve the records. Preserve the history.