Ep. 26: Genealogy + DNA | How to Use Artificial Intelligence Safely for DNA Research

If you have been sitting on a pile of DNA matches with no idea how to make sense of them, this episode was made for you.

Episode 26 of Ancestors and Algorithms tackles one of the most requested topics since the show launched: can you actually use artificial intelligence to help decode your DNA results without putting your family's genetic privacy at risk? The short answer is yes. But the HOW matters enormously, and most genealogists are either avoiding AI for DNA work out of fear, or diving in without understanding the real privacy risks. This episode fixes both problems.

In this episode, you will discover:

🧬 THE PRIVACY-FIRST TOOL GUIDE Five AI tools reviewed and rated for DNA safety. One completely free tool never trains on your uploaded data, ever. One popular tool requires a critical settings change before you use it for anything DNA-related. And one brand new health-specific workspace keeps your conversations completely isolated. You will know exactly what is safe, what needs configuring, and what you should never upload to any AI under any circumstances.

📊 5 DNA RESEARCH TASKS WHERE AI DELIVERS REAL RESULTS

• Centimorgan relationship analysis: what does 850 cM actually mean for your research?
• Shared match pattern decoding: figuring out which side of the family a match belongs to
• Understanding X-DNA inheritance, endogamy, and recombination in plain English
• Generating research hypotheses when your brick wall has you completely stumped
• Building an organized DNA research system that keeps all your matches and notes in one place

🔍 A REAL BRICK WALL SOLVED LIVE A 980 centimorgan match with zero shared surnames and completely different geographic origins. Two family trees going back six generations with absolutely nothing in common. One AI-generated research hypothesis changed everything and pointed to a Catholic institutional connection that neither tree had ever documented. You will hear the full story and the exact prompt that cracked it open.

💡 COPY AND PASTE PROMPTS INCLUDED Every technique comes with a ready-to-use prompt you can take directly into your own research today. No paid subscriptions required for any of it.

This episode is relevant whether your family roots are in the American South, New England, the British Isles, Ireland, Australia, Germany, Italy, Scandinavia, or Eastern Europe. DNA mysteries do not care about borders, and neither do the AI techniques covered here.

#GenealogyPodcast #DNAGenealogy #GeneticGenealogy #AIforGenealogy #FamilyHistory #AncestryDNA #23andMe #GenealogyTips #BrickWall #FamilyTree #AncestorsAndAlgorithms #DNAMatches #GenealogyResearch #FamilyHistoryResearch

Connect with Ancestors and Algorithms:

📧 Email: ancestorsandai@gmail.com
🌐 Website: https://ancestorsandai.com/
📘 Facebook Group: Ancestors and Algorithms: AI for Genealogy - www.facebook.com/groups/ancestorsandalgorithms/

Golden Rule Reminder: AI is your research assistant, not your researcher.

Join our Facebook group to share your AI genealogy breakthroughs, ask questions, and connect with fellow family historians who are embracing the future of genealogy research!

New episodes every Tuesday. Subscribe so you never miss the latest AI tools and techniques for family history research.

Transcript

So, here's the thing about DNA and genealogy that nobody really talks about. When DNA testing first exploded onto the genealogy scene back in the early 2000s, there was massive resistance from the traditional genealogy community. I mean, people were furious. This is cheating, they said. Real genealogists used documents, not spit in a tube. There were heated debates at conferences, angry blog posts, genealogy societies that refused to acknowledge DNA as legitimate research. Sound familiar? Because that's exactly what's happening right now with AI and genealogy. The same resistance, the same fear, the same accusations of cheating. But here's what I've learned after working with both DNA and AI for years now. The genealogists who embraced DNA early, they saw brick walls that had stumped researchers for generations. They found families torn apart by adoption. They identified unknown soldiers. They rewrote family histories. And I believe that genealogists who learned to safely integrate AI with their DNA research, they're about to do the same thing. Today, we're going to talk about something that makes a lot of genealogists nervous. Combining DNA analysis with artificial intelligence. And I'm going to tell you how to do it safely, ethically, and effectively. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host, Brian, and this might be the most important episode I've ever recorded. Let's dive in. 

Okay, before we go any further, I need to address the elephant in the room. I know a lot of you are thinking, wait, you want me to upload my DNA data to an AI chatbot? Are you out of your mind? And you know what? That's a completely legitimate concern. In fact, it's a smart concern. Your DNA data is some of the most personal, sensitive information you have. It reveals not just things about you, but about your parents, your children, your siblings, your cousins. People who never consented to having their genetic information analyzed by AI. So, let me be crystal clear about something right from the start. If you don't feel comfortable using AI tools for DNA analysis, I completely respect that decision. I am not here to pressure you, convince you, or tell you that you're wrong for being cautious. Privacy concerns about genetic data are valid, and you should never do anything with your DNA information that makes you uncomfortable. This episode is for the genealogists who have been wondering, could AI help me understand my DNA results better? Or, is there a safe way to do this? If that's not you, that's totally okay. You can skip this episode, or just listen to understand what others are doing. Still with me? Okay, good. Now, here's what I've discovered. There's a huge difference between uploading your raw DNA file to an AI chatbot and using AI to analyze information you derive from your DNA results. And that distinction is everything. Let me give you an example. I've got a DNA match on Ancestry. Let's call her Sarah. She shares 850 centimorgans with me. We have 47 shared matches. Our family trees overlap in one specific area, a small town in Pennsylvania in the 1870s. But I cannot for the life of me figure out exactly how we're related. Now, I could spend hours manually creating spreadsheets, drawing family trees, calculating possibilities. Or, I could describe this relationship puzzle to an AI tool and ask it to generate theories based on centimorgans ranges and generational distances. See the difference? In the first scenario, I'd be uploading Sarah's actual genetic data, her ethnicity breakdown, her chromosome browser information, potentially her raw DNA file. That would be a privacy violation. Sarah didn't consent to that. In the second scenario, I'm giving AI a mathematical puzzle. Two people share 850 centimorgans. They have common ancestors approximately four to five generations back. What are the possible relationship configurations? That's not genetic data. That's genealogical math. And that's the approach we're going to take in this episode. Now, I mentioned that DNA testing faced the same resistance that AI is facing now. And I think there's a lesson there. The genealogists who refused to use DNA, a lot of them eventually came around. Because they watched their colleagues solving mysteries they couldn't solve. They watched brick walls crumbling. They watched families reuniting. But here's the thing. The genealogists who succeeded with DNA weren't the ones who threw out traditional research methods. They were the ones who integrated DNA with solid documentary evidence. They used DNA as a tool, not a replacement. And that's our golden rule showing up again. AI is your research assistant, not your researcher. DNA results don't replace vital records. And AI doesn't replace your genealogical judgment. They're tools. Powerful tools, but tools nonetheless. So, here's what we're going to do today. First, I'm going to show you which AI tools are safest for DNA-related work and which ones you should avoid. Then, I'm going to walk you through five specific DNA genealogy tasks where AI can help without compromising anyone's privacy. And finally, I'm going to give you exact prompts and workflows you can use starting today. By the end of this episode, you're going to understand how to harness AI for DNA genealogy in a way that respects privacy, maintains ethical standards, and actually helps you solve real research problems. Sound good? Let's start with the tools. 

So, here's what I did before recording this episode. I spent an entire day researching the current privacy policies of every major AI tool. And I mean current as of February 2026 because this stuff changes fast. What I found surprised me. Some tools that I thought were safe, not so much. And some tools I've been avoiding actually have excellent privacy protections. Let me break this down for you because this is critical information. Tool number one, Notebook LM. This is going to sound like an ad, but I promise it's not. I do not get paid for these episodes. Google's Notebook LM is, hands down, the safest AI tool for analyzing DNA-related documents. Here's why. Notebook LM has a very simple, very clear privacy promise. It never trains its models on your uploaded data, ever. I verified this. I read their privacy policy. I checked multiple sources. And it's true. When you upload documents to Notebook LM, it analyzes them to answer your questions, but it doesn't use that information to improve its AI models. This is huge for DNA work. Here's how you'd use it. Let's say you downloaded your Ancestry DNA match list as a spreadsheet. You've got 500 matches, and you're trying to figure out which ones are related through your maternal grandmother's line versus your paternal grandfather's line. You can upload that spreadsheet to Notebook LM and ask it questions like, which matches share surnames from the Smith family in Ohio? Create a list of matches who have trees with ancestors in Pennsylvania between 1850 to 1900. Compare these matches and identify the common surnames that appear most frequently. Notebook LM will analyze your data, give you answers with citations, and then that's it. Your data stays your data. It's not training future AI models. It's not being used to improve Google's products. And the best part? It's completely free. Now, I need to be clear here. I'm not saying you should upload your raw DNA file to Notebook LM. That .csv file with your actual genetic markers? That stays on your computer. What I'm talking about is uploading derivative documents. Match lists, research notes, your analysis of relationships. Tool number two, Claude. Here's where things get a bit more complicated. Claude, made by Anthropic, used to be the gold standard for privacy. They promised they'd never train on user data. But in September 2025, they changed their policy for consumer accounts. Now, if you use Claude Free, Pro, or Max, you have the option to let them train on your conversations. It's opt-in, which means the default is that they don't train on your data. But you need to verify this in your settings. Here's what you do. Go to your Claude settings. Find the section about data usage. And make absolutely sure that training is turned off. Claude is still an excellent tool for DNA genealogy work. Why do I like Claude for this? Because it's really good at reasoning through complex relationship problems. It can take a scenario like person A shares 1,200 centimorgan with person B and person C shares 650 centimorgan with person A but only 425 centimorgan with person B. And help you figure out what that pattern means. But, and this is important, even with training turned off, I still wouldn't upload actual DNA files or chromosome browser screenshots with identifiable information. Use Claude for the mathematical and logical analysis, not for processing raw genetic data. Tool number three, ChatGPT Health. Now, this is interesting. In January, 2026, OpenAI launched something called ChatGPT Health. It's a separate, isolated space within ChatGPT specifically designed for health-related conversations. Here's what makes it different. Conversations in ChatGPT Health are encrypted separately, stored separately, and, this is a key part, not used to train OpenAI's foundation models. Could you use ChatGPT Health for DNA genealogy questions? Technically, yes. It's designed for health data, and genetic data falls under that umbrella. here's my take. I'd use it for understanding the science of genetic genealogy. Questions like, explain how XDNA inheritance works, or... What does it mean when siblings share different amounts of DNA with the same cousin, but not for analyzing your actual match data? For actual match analysis, I'd stick with notebook LM. Tool number four, perplexity. Perplexity is excellent for one specific DNA genealogy task, research. Let's say you're trying to understand endogamy in Ashkenazi Jewish communities, or you want to know more about how African American genealogists use DNA. A to trace pre-1870 ancestors, or you need to understand what phasing means in genetic genealogy. Perplexity excels at this because it searches the web and gives you answers with citations. You can ask, explain endogamy and how it affects DNA matching in genealogy research, and you'll get a comprehensive answer with links to genetic genealogy blogs, scientific papers, and educational resources. But perplexity isn't where you upload match data or analyze relationships. It's for learning, not for processing your specific data. Tool number five, Gemini, via Google AI Studio. One more tool worth mentioning. If you have old genetic genealogy documents, maybe a letter from a genetic genealogist from the early 2000s, or handwritten notes from a DNA conference, Gemini 3 through Google AI Studio has incredible handwriting transcription capabilities. But again, we're not uploading DNA results. We're using it as a transcription tool for genealogical documents that happen to be about DNA research. 

I'll cover it as a function of DNA research. Never upload to any AI tool Raw DNA Data Files, .csv, .txt from testing companies. Chromosome browser screenshots with match names visible. Full match list with email addresses or contact information. Genetic data about living people who haven't consented. Safe to work with in AI tools. Anonymized relationship questions. Like, two people share 850 Centimorgan. What's their likely relationship? Centimorgan math and probability calculations. General genetic genealogy education questions. Your own research notes about DNA names removed. Published genetic genealogy methodology questions. Make sense? I want to emphasize again, if any of this makes you uncomfortable, you don't have to do it. There are genealogists who solve DNA mysteries without ever using AI. And that's perfectly valid. AI is a tool, not a requirement. Use what works for you and respect your privacy boundaries. Alright, now that we've covered which tools to use, let's talk about how to use them. I'm going to walk you through five real DNA genealogy scenarios where I've used AI successfully. These are techniques I actually use in my own research. And I'm going to give you the exact prompts so you can try them yourself. Task number one, understanding Centimorgan ranges and relationship possibilities. This is probably the most common DNA question. Quote, we share X Centimorgans, how are we related? End quote. And here's where AI is actually brilliant. Because it can instantly calculate all the possible relationships that fit a specific Centimorgan range. Let me show you how I do this. Let's say I have a match who shares 425 Centimorgans with me. We have no shared surnames in our trees. I need to figure out how we're related. I open Notebook LM because it's private and won't train on this. And I use this prompt. Based on genetic genealogy standards, what are the possible relationships between two people who share 425 Centimorgans? Please provide, one, the most likely relationships. Two, the possible generational distance. Three, the probability of each relationship type. Four, what additional information would help narrow this down? Notebook LM comes back with a comprehensive answer. 425 Centimorgan could be a great-grandparent, great-grandchild, great-aunt, great-uncle, half-aunt, half-uncle, first-cousin, or half-niece, half-nephew. But here's where it gets really useful. I can then ask follow-up questions. If this match is 425 Centimorgan and we have no shared surnames in our family trees going back to 1800, what does that suggest about the relationship? What are the most likely scenarios? And now AI helps me think through possibilities I might not have considered on my own. Could be a relationship through a female line where surnames changed. Could indicate an adoption or name change. Might suggest the connection is through a less-documented immigrant ancestor. Could point to a non-parental event, NPE, situation. See what happened there? I didn't give AI anyone's actual genetic data. I gave it a math problem and asked it to help me think through logical scenarios. This is AI as your research assistant, not your researcher. I still have to go verify everything with actual genealogical research. But AI helped me brainstorm possibilities I should investigate. Task number two, analyzing shared match patterns. Here's a more complex scenario. Let's say you're working with three DNA matches. Match A, 650 centimorgan to you. Match B, 480 centimorgan to you. Match C, 520 centimorgan to you. And here's the interesting part. Match A and Match B share 920 centimorgan with each other. Match A and Match C share 725 centimorgan with each other. Match B and Match C share 0 centimorgan with each other. What does that pattern tell you? This is where AI really shines because it can process these complex relationship triangulations faster than you can draw them out on paper. Prompt for Claude with training off. I need help understanding a DNA match pattern. I'm trying to determine how three matches are related to me. Match A shares 650 centimorgan with me. Match B shares 480 centimorgan with me. Match C shares 520 centimorgan with me. Match A and Match B share 920 with each other. Match A and Match C share 725 centimorgan with each other. Match B and Match C share 0 centimorgan with each other. What does this pattern suggest about? 1. Which side of my family these matches come from? 2. How Match A is likely related to Matches B and C? 3. What this tells me about my research strategy? Claude's analysis would likely point out that since B and C don't match each other at all, they're definitely from different sides of your family. Or different sides of Match A's family. This is the kind of pattern recognition that takes humans a long time to work out. But AI can spot instantly. Again, notice what I didn't do. I didn't upload anyone's family tree. I didn't share names. I didn't provide ethnicity data. I gave AI a logic puzzle using numbers. Task number 3. Understanding Complex Inheritance Patterns. Okay, this one gets technical, but it's where AI is absolutely invaluable for education. Let's say you're trying to understand why you and your sibling have different amounts of DNA from the same grandparent. Or you're confused about ex-DNA inheritance and why it matters for genetic genealogy. These are complex scientific concepts that traditional genetic genealogy books explain in dense academic language. But AI can break them down in ways that make sense. Prompt for Perplexity. Explain ex-DNA inheritance patterns in genetic genealogy in simple terms. Include how ex-DNA differs from autosomal DNA. Which ancestors can pass ex-DNA to you? Why this matters for genealogy research? Common misconceptions about ex-DNA. Please provide sources from reputable genetic genealogy resources. Perplexity will search actual genetic genealogy blogs, research papers, and educational sites, then synthesize that information with citations. The beauty of this approach? You're not relying on AI to generate information about genetics, which could be wrong. You're using it to find and synthesize information from trusted sources. Remember our golden rule. In this case, it's acting like a research librarian who knows where to find the best explanation of complex topics. Task number 4. Generating Research Hypothesis for Brick Walls. This is one of my favorite uses of AI for DNA work. Let's say you've hit a brick wall with a DNA mystery. You have matches, you have data, but you can't figure out the connection. AI can help you generate research hypothesis? Possible explanations you should investigate. Here's a real scenario I worked on. match. I had a DNA match who shared 980 centimorgans with me. That's a significant amount. Likely a first cousin or great aunt, great uncle range. But there were no shared surroundings between our trees going back six generations. I created a research summary document, names removed for privacy, and uploaded it to Notebook LM. Document summary. DNA match information. Shared centimorgan, 980. Largest segment, 142 centimorgan. Total shared segment, 34. Shared matches, 127. Age difference, approximately 50 years. Matches older. Tree analysis. My tree. Documented back to 1750 in Pennsylvania, Ohio, Germany. D. A. Pages tree. Documented back to 1780 in New York, Vermont, Ireland. No surname overlap. No geographic overlap in direct lines. Both families' Catholic background. Research question. How can two people share 980 centimorgan with no apparent family connection? Then I asked Notebook LM. Prompt. Based on this DNA match information, generate 10 possible research hypotheses that could explain this relationship. Consider possible adoptions or name changes. Female lines that might connect us. Immigration patterns. Religious-ethnic community connections. Documentation gaps. For each hypothesis, suggest what records I should search to prove or disprove it. And here's what happened. Notebook LM generated hypotheses I hadn't considered. One of them was, consider whether either family had connection to Catholic orphanages or charitable institutions. Which were common in both Pennsylvania and New York in the mid-1800s. That single hypothesis led me to research Catholic orphanages in both states. And guess what? I found a connection through a great-great-aunt who had been placed with a Catholic charity in Philadelphia, then adopted by a family in New York. Would I have thought of that on my own? Maybe eventually. But AI helped me get there faster by systematically generating possibilities based on the data I provided. Task number five, creating DNA research study guides. Here's a use case that I think will blow your mind. Using AI to create customized study materials for understanding your own DNA results. Let's say you've been working on a specific DNA mystery for months. You have notes scattered everywhere, in notebooks, in ancestry message threads, in Word documents, in emails to other researchers. You can upload all those notes with names anonymized to NotebookLM and ask it to... Prompt Based on all the documents I've uploaded about this DNA research problem, create 1. A timeline of what I've discovered so far 2. A summary of the key questions still unanswered 3. A list of the matches involved in their relationships to each other 4. A research plan for the next three months with specific tests 5. A one-page cheat sheet I can reference when working on this mystery NotebookLM will analyze all your scattered research and synthesize it into organized, actionable information This is particularly useful if you're working on a complex NPE, non-parental case or an adoptee search where you have dozens of matches and need to keep track of how everyone relates to everyone else Before we move on, let me give you one more reminder. Another thing I just showed you involved using AI to analyze patterns, generate hypotheses and organized information Not to process raw genetic data If at any point you felt uncomfortable with these techniques, that's valid DNA research is deeply personal. Some people prefer to work through these mysteries entirely on their own or with human genetic genealogists So, you may not be surprised, but you might have been genetic genealogists There's no wrong choice here The goal of this episode isn't to convince everyone to use AI for DNA work It's to show those who are interested that there are SAFE, ethical ways to do it

All right, let's bring this home with some practical guidelines and the homework assignment And so, we've covered the tools and the techniques Now I want to give you a framework for making ethical decisions about when and how to use AI with DNA data I call this the DNA plus AI decision tree and it goes like this Question 1: Does this involve raw genetic data? Yes, don't use AI, period No? Move to question 2 Question 2: Does this involve identifiable information about living people who haven't consented? Yes, anonymize first, or don't use AI No? Move to question 3 Question 3: Could this analysis be done without AI in a reasonable amount of time? Yes, consider doing it manually first No, AI can be a helpful tool Question 4: Am I using a privacy respecting AI tool? No? Switch to notebook LM or another safe option? Move to question 5 Question 5: Am I treating AI as an assistant, not a researcher? No? Rethink your approach Yes, you're good to proceed This framework keeps you on the ethical side of DNA plus AI work And here's the most important principle underlying all of this When in doubt, protect privacy first If you're not sure whether something is safe to upload to an AI tool, don't upload it If you're uncertain whether you have someone's consent, assume you don't If you can't anonymize data effectively, don't use AI for that task Now, one more thing I want to address The question of whether AI will "replace human genetic genealogies" And the answer is absolutely not Here's why Genetic genealogy requires not just mathematical analysis, but cultural understanding, historical context, ethical judgment, and emotional intelligence When someone discovers through DNA that their grandfather wasn't who they thought, that's not a math problem That's a human situation that requires compassion, wisdom, and experience AI can help you to calculate the symptom organ probabilities It can not help you navigate the emotional complexity of an NPE discovery AI can generate research hypotheses It can not interview elderly relatives with sensitivity and care AI can organize your match data It can not reach out to a DNA match and build a relationship with them AI is your research assistant, not your researcher And it's definitely not your genetic counselor, your therapist, or your family mediator The genealogists who thrive in this new era are the ones who use AI as a tool to enhance their human skills Not replace them Alright, let's wrap this up with your homework assignments Here's what I want you to do this week Option #1: The Curious Beginner If you're new to this whole DNA plus AI thing, start small Number one, go to Notebook LM. It's free, remember. 2. Ask it this question, quote, explain the difference between autosomal DNA, X-DNA, Y-DNA, and mitochondrial DNA in terms a genealogy beginner would understand, end quote. 3. Read the answer, ask follow-up questions. 4. Get comfortable with how Notebook LM works. That's it, no uploading DNA data, just practicing with educational questions. Option 2, the active DNA researcher. If you're already working with DNA matches, try this. 1. Pick one DNA match you've been puzzled by. 2. Write down, anonymously, the shared Centimorgan amount, the generational distance in your trees, and any patterns you've noticed. 3. Upload that information to Notebook LM. 4. Ask it to help you generate five research hypotheses. 5. Pick one hypothesis and spend a week testing it with traditional genealogical research. 6. Option 3, the advanced challenge. 7. For those of you ready to dive deep. 7. Take a complex DNA mystery you've been working on. 8. Gather all your research notes, anonymize names if needed. 9. Upload them to Notebook LM. 10. Ask it to create a comprehensive research summary and action plan. Number 5. Use that AI-generated plan to guide your next month of research. 

Thank you so much for listening to Ancestors and Algorithms. I know this episode might have pushed some comfort zones, and that's okay. The goal wasn't to convince everyone to use AI for DNA research. It was to show that for those who are interested, there are safe ethical ways to do it. If you found this helpful, please share it with another genealogist who's been curious about this topic. And join our Facebook group, Ancestors and Algorithms, AI for Genealogy. Where we can discuss these techniques and share our experiences. Before I let you go, I have some exciting news. I'll be attending RootsTech 2026 in person. For those of you who don't know, RootsTech is the world's largest genealogy conference, and this year it runs March 5th through the 7th in Salt Lake City. I am beyond excited to be there, soaking up everything I can about the latest in AI and genealogy research, and you know I'm going to bring it all back here to share it with you. If you're going to be there in person, I would genuinely love to meet you. Come find me. Say hello. Tell me about your brick walls. I want to hear it all. Let's turn this community from a podcast into a real-life reunion. RootsTech 2026, March 5th through 7th. I hope to see some of you there. I'm your host, Brian, and I will see you next week for another journey into the past, powered by the future. Until then, happy researching.