On a Friday afternoon in late April 2018, a retired cold-case investigator named Paul Holes sat in his office in Contra Costa County, California, and stared at a printout of distant cousins. The cousins did not know each other. They did not know they were related. They did not know that the man whose DNA had brought them all together on a free genealogy website had also raped at least fifty women and murdered at least thirteen people between 1973 and 1986.

Four days later, on April 24, 2018, the Sacramento County Sheriff arrested seventy-two-year-old Joseph James DeAngelo — a former police officer and Vietnam veteran — in the driveway of his suburban home. DeAngelo was the Golden State Killer. He was the East Area Rapist. He was the Original Night Stalker. For forty-two years, his name had been the most-wanted blank in California law enforcement. A genealogy website had filled it in.

What Holes and the genealogist Barbara Rae-Venter had just done would, within four years, become the most disruptive forensic tool of the twenty-first century. It is called forensic genetic genealogy. It is responsible, as of 2026, for the resolution of more than six hundred cold cases that investigators had effectively abandoned. And it works on a logic that nobody had really put together before that week in 2018.

The shape of a stranger’s family tree

Traditional forensic DNA, the kind we have used in courtrooms since the early 1990s, works on a closed comparison. Investigators take DNA from a crime scene, generate a short profile from about twenty genetic markers, and compare it against a federal database called CODIS, which holds the DNA profiles of about twenty-one million people — mostly convicted felons, arrestees, and missing persons. If the killer is in CODIS, you get a hit. If they are not, you get nothing. For decades, the great frustration of cold-case investigators was that most violent offenders are not in CODIS. Many have never been arrested. Many committed their crimes before the database existed.

Forensic genetic genealogy works on a completely different principle. Instead of looking for an exact match, it looks for the killer’s relatives.

The same crime-scene DNA is processed into a much richer profile — hundreds of thousands of genetic markers known as single-nucleotide polymorphisms, or SNPs. That SNP profile is then uploaded to public consumer genealogy databases like GEDmatch and FamilyTreeDNA. Those databases hold the DNA of millions of recreational genealogists — people who took a 23andMe or AncestryDNA kit and uploaded their results to find lost cousins, fill in family trees, or track immigration histories.

The killer almost certainly never uploaded their own DNA. But statistically, in a database of millions of Americans of European descent, the killer’s third or fourth cousins almost certainly did. And if you have several of those cousins, a competent genealogist can work backwards through marriage records, census data, and obituaries to narrow that web of relationships down to a single person.

The case that proved the method

The Golden State Killer case is the one everyone knows. It was the proof of concept. It was also, by no coincidence, the case that had haunted Paul Holes longer than any other in his career. He had been chasing the killer’s DNA for more than two decades.

In late 2017, Holes connected with Barbara Rae-Venter, a retired patent attorney who had taken up genealogy as a hobby and become one of the most skilled SNP-comparison genealogists in the country. They uploaded the killer’s profile to GEDmatch in April 2018. They got a handful of third- and fourth-cousin hits. From those hits, Rae-Venter built two family trees — one for each side of the killer’s family. The trees converged on a single great-great-great-grandfather who had lived in Sacramento in the 1800s. From there, they worked forward through marriages and births until they identified a small set of male descendants who would have been the right age and in the right place to be the killer. Joseph James DeAngelo was on that list. Surveillance officers retrieved a discarded tissue from his garbage. The match was confirmed in four days.

DeAngelo pleaded guilty in June 2020 to thirteen counts of first-degree murder and admitted to dozens of additional rapes that could no longer be prosecuted because the statute of limitations had run out. He is serving multiple consecutive life sentences without parole. He is now in his eighties.

The cases that followed

The Golden State Killer arrest was the public spark. The fire spread fast. Within a year, dozens of small police departments around the country were partnering with private genealogy labs to apply the same method to their own cold cases.

Within weeks of DeAngelo’s arrest, the murder of twelve-year-old April Tinsley in Fort Wayne, Indiana — a case that had gone cold in 1988 — was solved using the same technique. The killer, John Miller, had taunted police for years with handwritten notes, even calling the Tinsley family directly. He confessed within hours of being arrested.

In 2019, William Earl Talbott II was convicted of the 1987 double murder of Jay Cook and Tanya Van Cuylenborg, a Canadian couple killed during a trip to Seattle. It was the first forensic-genealogy case to go to trial, and the conviction stuck.

In December 2022, the Philadelphia Police Department announced that the boy whose body had been found beaten and abandoned in a JC Penney bassinet box in 1957 — a case known for sixty-five years only as “America’s Unknown Child” or “The Boy in the Box” — had finally been identified through forensic genealogy as four-year-old Joseph Augustus Zarelli. His murderer remains unknown, but for the first time in nearly seven decades, the child had his name back.

By the end of 2025, the technique had also helped identify the long-unknown victims of serial killers such as John Wayne Gacy and the Green River Killer; cleared the names of several people wrongly convicted on faulty hair-comparison evidence; and resolved decades-old missing-persons cases ranging from a 1979 unidentified murder victim in upstate New York to a Jane Doe found floating in the Salt River in 1989. The total count, tracked by the nonprofit DNA Doe Project and similar groups, sits north of six hundred cases and climbs every month.

The labs that made it scalable

The Golden State Killer case was solved with a very good sample of crime-scene DNA. Most cold cases do not have that luxury. The blood is decades old. The semen has degraded. The bone is fragmentary. CODIS-style testing can sometimes generate a partial profile from these samples, but the SNP-density required for genealogy work was, for a long time, considered impossible to extract from heavily degraded material.

That has changed because of a small number of specialized private labs. The most prominent is Othram, a Texas company founded in 2018, which developed a process called forensic-grade genome sequencing. It can build a usable SNP profile from samples that traditional labs would dismiss as unworkable — a few microscopic skin cells, a fragment of hair, a partial bone shard. Parabon NanoLabs, in Virginia, and the DNA Doe Project, a volunteer-driven nonprofit, have also contributed hundreds of resolutions.

Most of these labs operate on a fee-for-service basis. A genealogy case typically costs a police department between $7,000 and $15,000 — a fraction of the cost of a single homicide detective’s year of work, and orders of magnitude cheaper than a wrongful-conviction lawsuit. Crowdfunding platforms have sprung up specifically to pay for genealogy testing on cases where the local jurisdiction cannot afford it.

The ethical fault line

The technique has critics, and the criticism is serious.

When a recreational genealogist uploads their DNA to GEDmatch to find their long-lost half-sister, they are also — mathematically — making every distant cousin they have ever had identifiable to law enforcement. They are not just exposing themselves. They are exposing several hundred relatives, most of whom have never consented to being in any kind of investigative database.

In May 2019, GEDmatch — then the workhorse database for genealogy investigations — changed its terms to require users to actively opt in to law-enforcement matching. The percentage of opted-in profiles fell sharply, and the technique’s effectiveness took a noticeable hit. FamilyTreeDNA has taken a different approach: it cooperates with law enforcement requests by default, but only for violent felonies and unidentified human remains. 23andMe and AncestryDNA, the two largest commercial DNA testing companies, do not allow law enforcement access at all without a court order, which is exceedingly rare to obtain.

Courts have largely upheld the practice. A 2023 federal court ruling in the case of State v. Hartman held that uploading a SNP profile to a public database, where users have notice that their data may be searched, does not constitute a Fourth Amendment search. The U.S. Department of Justice issued interim guidelines in 2019 (still in force, with minor amendments through 2024) restricting federal use of the technique to violent crimes and unidentified-remains cases, and requiring written approval from a senior prosecutor before each search.

The deeper question — what privacy means in a world where your second cousin can give away your future privacy by mailing a tube of spit to a Utah startup — remains genuinely open. Genealogists and civil-liberties lawyers have started using the phrase genetic informational consent to describe a kind of consent that almost no one in America has actually been asked to give.

What it means for the cases that are still open

For families of the murdered and the missing, the calculus is much simpler. A grandmother in Wisconsin whose daughter vanished in 1992 does not particularly care about the philosophical debate over genetic informational consent. She wants to know what happened. For a growing number of those grandmothers, the answer is now arriving — sometimes after thirty or forty years of silence, sometimes by a phone call out of nowhere from a detective who says: We know who he is.

The technique’s biggest limitation is also its strangest strength. It works best on samples whose owner has many genealogical relatives in the consumer databases — which means it works disproportionately well on people of European descent and disproportionately poorly on Black, Hispanic, and Asian Americans, who are systematically underrepresented in genealogy databases. Researchers and database operators are working to close that gap, but the gap is real, and it shapes which cases get solved first.

Eight years in, forensic genetic genealogy is no longer a novelty. It is a standard step in any well-funded cold-case investigation. The FBI, most state crime labs, and a majority of large municipal police departments now have at least one investigator who knows how to build a SNP profile, write a search request, and read a family tree.

It is the closest thing American policing has had to a true revolution since the introduction of DNA fingerprinting in 1986. And it began, fittingly, with a retired investigator, a retired patent attorney, a free website, and a man who had thought — for forty-two years — that he had gotten away with everything.

---

Sources: U.S. Department of Justice interim policy on Forensic Genetic Genealogical DNA Analysis and Searching (September 2019, amended 2024); DNA Doe Project public case log; Othram Inc. case database; Paul Holes, Unmasked: My Life Solving America’s Cold Cases (2022); Barbara Rae-Venter, I Know Who You Are (2023); Center for Genetics and Society policy briefs.