Longeye: Police Detectives Have a New AI Partner. Genealogists Need One Too.
An AI tool called Longeye is helping law enforcement crack cases by analyzing mountains of digital evidence in minutes. Here are three reasons a similar tool would be a game-changer for family history research — and why genealogists should be paying close attention.
If you’ve spent any time in genealogy, you already know the feeling: you’re staring at a screen full of census records, ship manifests, church registers, and scanned letters — and somewhere in that pile is the connection that ties your 3x great-grandmother to a specific parish in County Cork. You know it’s there. You just need to find it.
Now imagine an AI tool that could ingest all of that evidence, index it, cross-reference it, and point you toward the patterns you’re missing — without pulling in random data from the internet, without hallucinating ancestors, and without telling you what to think.
That tool doesn’t exist for genealogists yet. But something very close to it already exists for police detectives. And it’s working.
What Is Longeye — And Why Should You Care?
Longeye is a San Francisco-based AI platform built specifically for law enforcement investigations. As reported by The Washington Post in April 2026, approximately 35 law enforcement agencies at the local, state, and federal level are in the early stages of adopting the tool to process the overwhelming volume of digital evidence that comes with modern casework.
The concept is straightforward but powerful: detectives upload their case evidence into a secure, closed environment. The AI then analyzes, indexes, and cross-references everything — documents, audio, video, data — surfacing patterns, connections, and relevant details that a human investigator might take weeks or months to find manually.
The results have been striking. Detective Lauren Cunningham of the Oklahoma City Police Department told the Post that the 20 or so hours per week she’d normally spend monitoring jail calls from murder suspects dropped to fewer than five. A colleague was able to sift through thousands of pages of financial documents and spot patterns in a fraction of the usual time. And a sex crimes investigator used Longeye to translate suspect phone calls into English — uncovering a confession that turned a child rape case headed for trial into a likely plea agreement.
The scale is significant: CEO Guillaume Delépine told the Post that so far in 2026, Longeye has compressed roughly 34 years’ worth of detective work into just a few months, processing 25 million files across those 35 agencies. Oklahoma’s Department of Corrections is also testing the tool to analyze nearly 50 million minutes of annual prison phone calls and data from thousands of confiscated contraband cellphones.
But here’s the detail that matters most for our purposes: Longeye doesn’t create information, doesn’t predict outcomes, and doesn’t pull in data from outside the case files. It operates in what Delépine calls a “closed sandbox” — entirely walled off from outside information that could compromise the analysis. Each case file is self-contained, pulling only from what investigators have uploaded. Every finding generates a citation linking back to the original source material, preserving a clear chain of custody.
Sound familiar?
“I didn’t know I was ever going to work in a place where AI understood that its sole job was to be a fact finder.” — Detective Lauren Cunningham, Oklahoma City Police Department
Genealogists Are Detectives. Full Stop.
I’ve said this for years, and I don’t think it’s an exaggeration: genealogists are investigators. We gather evidence. We evaluate the reliability of sources. We follow leads, test hypotheses, work through contradictions, and build cases that either hold up to scrutiny or send us back to the drawing board.
The parallel with criminal investigation isn’t a stretch — it’s structural. A detective working a case and a genealogist working a research question face remarkably similar challenges: too much data, not enough time, scattered sources across multiple formats, and the constant risk that a single overlooked detail changes the entire picture.
If an AI tool designed for one kind of investigator is producing measurable results, then we should absolutely be asking: what would a similar tool do for the other kind?
I think the answer is a lot. And I want to walk through three specific capabilities that make Longeye’s approach so well-suited to genealogy research — with the critical caveat that I’ll repeat throughout this piece: AI is an assistant, not a substitute for the genealogist’s judgment.
Three Ways a Longeye-Type AI Would Transform Genealogy Research
1. Walling Off Outside Data and Influences
This is the feature that made me sit up straight.
One of Longeye’s core design principles is isolation — what the Post describes as a “closed sandbox.” Each investigation is self-contained. The AI doesn’t reach out to external databases, doesn’t cross-reference with the open internet, and doesn’t train on data from other cases. It analyzes only what the investigator has uploaded. Period. As Oklahoma’s corrections chief of operations Jason Sparks put it to the Post, law enforcement needs a tool that’s self-contained — one where outside noise doesn’t contaminate the analysis.
For genealogists, this kind of closed-environment AI would be transformative — and here’s why.
Right now, when you use a general-purpose AI chatbot for genealogy help, you’re getting responses drawn from its entire training corpus. That includes every unsourced family tree on the internet, every Find A Grave memorial with an unverified date, every Ancestry hint that somebody blindly accepted fifteen years ago. The AI doesn’t distinguish between a carefully documented proof argument and a Rootsweb page from 2003 that attributes twelve children to a woman who died at age twenty-four.
A walled-off AI genealogy tool would eliminate that noise entirely. You would be working with your evidence — the documents, records, and sources that you have vetted. The AI wouldn’t be tempted to fill gaps with data it scraped from someone else’s tree. It would tell you what your evidence shows, flag what it doesn’t, and let you decide what to do about it.
This isn’t about limiting the AI’s power. It’s about focusing it. A detective doesn’t want the AI freelancing with evidence from unrelated cases, and a genealogist doesn’t want the AI polluting a carefully built research file with data from the wilds of the internet. The discipline of working within a defined evidence set is the whole point.
2. Uploading Your Existing Research Documents and Images
Here’s the practical heart of the concept: you upload what you’ve got.
Longeye accepts virtually any digital format — audio, video, images, documents, data. According to the Post, each case file entered into the system can be turned into timelines, maps, or spreadsheets, and the AI can identify key moments from hundreds of hours of witness interviews or phone calls. When Detective Cunningham uploads her case materials, the system processes them and provides what she calls “insights” and “reference points” — not conclusions, but leads worth investigating.
Now think about the average genealogist’s research collection: scanned census records, digitized birth and death certificates, church register images, handwritten letters, newspaper clippings, DNA results, research logs, probate documents, land deeds, pension files. Many of us have hundreds — even thousands — of files scattered across folders, drives, and filing cabinets.
An AI genealogy assistant that could ingest all of that material and make it searchable and cross-referenceable in seconds would be extraordinary. Imagine uploading a folder of fifty scanned documents related to a single family line and asking the AI to identify every mention of a particular name, every date, every location — then map those data points against each other for patterns and contradictions.
That’s not science fiction. That’s what Longeye is already doing for police investigators, just in a different domain.
A practical example: You’ve got a pension file, three census records, a church baptism register, and a county history book page — all referencing your ancestor. The documents contain slightly different birth years and name spellings. A Longeye-type AI could review every instance of that ancestor across all uploaded files, lay out the discrepancies side by side, and help you focus your analysis on the points of conflict. You still make the call. The AI just makes sure you’re seeing everything.
And here’s where I want to emphasize the assistant framing again: the AI doesn’t tell you which birth year is correct. It doesn’t “resolve” the conflict for you. It presents your evidence clearly and completely so that you can apply your knowledge of the records, the time period, the informant reliability, and the historical context to reach a reasoned conclusion. The thinking is still yours. The AI is your research assistant, not your research director.
3. Accelerating The Genealogical Proof Standard
This is where the genealogy-specific potential of a tool like this gets really exciting.
For those unfamiliar, the Genealogical Proof Standard (GPS) is the framework developed by the Board for Certification of Genealogists® that defines what constitutes credible genealogical research. It has five elements:
The Five Elements of the Genealogical Proof Standard
- Reasonably exhaustive research has been conducted
- Each statement of fact has a complete and accurate source citation
- The evidence has been reliably analyzed and correlated
- Any contradictory evidence has been resolved
- The conclusion has been soundly reasoned and coherently written
The GPS is the gold standard for a reason: it’s what separates documented family history from casual guesswork. But let’s be honest — rigorously applying all five elements is time-consuming, especially for genealogists who aren’t writing for publication or certification. Many hobbyist researchers understand the GPS in principle but struggle to apply it consistently in practice.
A Longeye-type AI tool could serve as a GPS accelerator. Let me walk through how it could assist with each element — while keeping the genealogist firmly in the driver’s seat.
- Reasonably exhaustive research. The AI could review your uploaded evidence and flag gaps — “You have census records for 1850 and 1870, but nothing for 1860. You have a marriage record but no church record for the same event.” It doesn’t conduct the research for you, but it helps you see where your coverage is thin so you know where to look next.
- Complete and accurate source citations. Here’s where AI can provide enormous practical value. Formatting proper source citations following Elizabeth Shown Mills’ Evidence Explained is something many genealogists find tedious and error-prone. An AI trained on citation standards could take a document you’ve uploaded — say, a digitized probate record — and generate a properly formatted citation that you can review, adjust, and adopt. That’s not replacing your judgment; it’s handling the formatting busywork so you can focus on the analysis.
- Analysis and correlation of evidence. This is where cross-referencing power shines. Given a set of uploaded documents, the AI could identify where multiple sources agree, where they conflict, and where information in one record illuminates something unclear in another. It could produce a summary matrix of how your evidence items relate to each other — the kind of evidence-correlation table that professional genealogists build by hand.
- Resolution of conflicting evidence. The AI won’t resolve the conflict for you — that requires human judgment, historical knowledge, and critical thinking. But it can make sure you’ve seen every conflict in your evidence set. One of the most common GPS failures is not that genealogists reach the wrong conclusion, but that they never notice the contradictory evidence in the first place. An AI that systematically surfaces every discrepancy across your documents addresses that problem directly.
- Soundly reasoned, coherently written conclusion. After you’ve done the analysis and reached your conclusion, the AI could assist with the writing — helping you draft a proof summary or proof argument that lays out the evidence, addresses the conflicts, and presents your reasoning clearly. Again: your reasoning, your conclusion. The AI helps you articulate it in a structured, standards-compliant format.
The AI doesn’t make you a better thinker. It makes sure you’re seeing all of your evidence — and helps you present your thinking more clearly.
Why “AI as Assistant” Isn’t a Disclaimer — It’s the Whole Point
I want to be direct about something, because I think it matters for how our community approaches these tools.
What struck me most about the Post piece wasn’t the technology — it was Detective Cunningham’s skepticism before she used it. She’d seen the ways AI was making the world worse: hallucinations, misinformation, manipulation. She was not an easy sell. What won her over was that Longeye was designed to be a fact finder, not an answer generator. It was firewalled from the public, federally compliant, and entirely reliant on her original detective work.
That distinction matters enormously. Marc Caudel, a private criminal defense investigator who has worked with innocence organizations and federal public defenders, told the Post he sees AI as a way to level the playing field — comparing it to employing an intern who can cut through the mundane labor of reviewing discovery to find the details that matter. Not a replacement for the investigator. A force multiplier.
That is exactly the right model for genealogy AI.
We don’t need an AI that builds family trees for us. We don’t need one that “solves” our brick walls by inventing plausible-sounding connections. We’ve all seen what happens when people accept hints and matches without critical evaluation — it’s how we end up with trees that have Civil War veterans born in 1750 and women bearing children at age eight.
What we need is an AI that makes us more efficient at the work we already do well: gathering evidence, organizing it, spotting patterns, identifying gaps, and documenting our findings. An AI that respects the process rather than trying to shortcut it. Delépine told the Post that building this kind of tool right means building a “much more deeply thinking machine.” That’s the bar genealogists should set for any AI tool we adopt.
Longeye is proof that this kind of tool can work in a high-stakes investigative context where accuracy and source verification are non-negotiable. The genealogy community should be paying attention — and, frankly, we should be vocal about what we’d want in a genealogy-specific version.
The Bottom Line
Longeye wasn’t built for genealogists. But the problems it solves — too much evidence to process manually, details buried in large document sets, the need for every finding to link back to a verifiable source — are our problems too.
A genealogy AI tool that lets you upload your own curated research files, works exclusively within that closed evidence set, and helps you meet the Genealogical Proof Standard more efficiently? That’s not hype. That’s the tool I’ve been waiting for.
If someone builds it, I’ll be first in line. And I suspect I won’t be alone.
Have thoughts on AI tools for genealogy? Drop me a line at hidefgen@gmail.com or leave a comment — I’d love to hear what features you’d want in a genealogy-specific AI research assistant.
* * *
Author’s Note: I want to be transparent that this content – Longeye: Police Detectives Have a New AI Partner. Genealogists Need One Too. – was created in part with the help of an artificial intelligence (AI) language model – Claude Opus 4.6. The AI assisted in generating an early draft of the content, but every paragraph was subsequently reviewed, edited, and refined by me. The final content is the result of extensive human curation and creativity. I am proud to present this work and assure readers that while AI was a tool in the process, the story, style, and substance have been carefully shaped by the author.
