Ep. 37: What Grandma Never Said - Using AI to Uncover Hidden Ancestors in Census Records

Every family has a story that got quietly handed down across the generations. A birthplace. A number of children. One marriage, one life, neatly summarized. But what happens when you sit down with the actual records and the story doesn't match?

In this episode of Ancestors and Algorithms, Brian walks through one of the most universal genealogy research scenarios there is: testing a family oral history against primary documents using four free AI tools. What starts as a simple census comparison becomes the discovery of a hidden first marriage, a child no one in the family ever mentioned, and a woman who rebuilt her life in silence after tragedy.

If you have ever accepted a piece of your family story at face value, this episode is for you.

In this episode, you will learn:

• How to use Claude AI to build a cross-census comparison table that surfaces inconsistencies your eyes might miss
• How to use ChatGPT to generate a targeted research checklist for finding a missing marriage or undocumented children
• How to use Perplexity to verify which genealogy records actually exist for your ancestor's state and time period before you waste hours searching in the wrong place
• How to use NotebookLM to organize all your gathered evidence, build a timeline from your uploaded documents, and identify the specific gaps that still need to be filled
• What the "children born" and "children living" columns in the 1910 federal census actually reveal, and why most researchers walk right past them
• How to recognize a second marriage in a census record and what records to search next

The AI tools featured in this episode (all free tiers):

• Claude by Anthropic (claude.ai)
• ChatGPT by OpenAI (chatgpt.com)
• Perplexity (perplexity.ai)
• NotebookLM by Google (notebooklm.google.com)

Records and resources mentioned:

• FamilySearch Indiana Marriages 1811-2019 (free at familysearch.org)
• Indiana State Archives vital records guidance (in.gov/iara)
• 1900, 1910, 1920, and 1930 US Federal Census (free at familysearch.org and ancestry.com)
• Hoosier State Chronicles Indiana newspapers (free at newspapers.library.in.gov)
• Chronicling America historic newspapers (free at chroniclingamerica.loc.gov)

For Australian and New Zealand researchers: The techniques in this episode translate directly to your family history research. Use electoral rolls on the National Archives of Australia website (naa.gov.au) as a census substitute for early 20th century ancestors. State Births, Deaths, and Marriages registries hold marriage records that can surface a first marriage the family never mentioned.

For UK and Irish researchers: England and Wales civil registration indexes marriages from 1837. FreeBMD at freebmd.org.uk gives you free access to birth, marriage, and death indexes going back to that date. Scotland's records are searchable at ScotlandsPeople (scotlandspeople.gov.uk). The family story that nobody told exists in British and Irish families exactly as it does in American ones.

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

SPEAKER_00 0:12

There's a moment in genealogy research that every one of us knows. You think you're filling in a date, maybe a birth year, maybe a marriage year. You've heard the story a hundred times. You know how it goes. And then you look at the actual record. I was looking at a census image nineteen ten. A woman I'll call Edna. The family had always said three children, one husband born in Ohio. Standard family story. Nothing complicated. I was confirming what we thought we already knew. And then I saw two numbers that changed everything. Children born five. Children living four. And next to her marital status, M2, second marriage? Nobody in this family had ever mentioned a first marriage. Nobody had ever mentioned five children. Nobody had ever said a single word about whatever happened to the children who weren't in that family tree. Today we're going to follow that thread. And I want you to follow along carefully because I promise you there is a version of this story sitting somewhere in your own family tree right now. The story that nobody told you. The chapter that got quietly closed. We're going to use four AI tools to open it back up. Let's dive in. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host, Brian, and today we're doing something that I think is going to resonate with almost every single person listening to this episode. We're talking about the family story that turned out to be wrong. Not maliciously wrong, not dramatically wrong, just quietly, tenderly, humanly wrong in ways that a stack of census records can reveal with startling precision. Now, if you've been listening for a while, you know we spend a lot of time on ancestors from the mid-1800s. Immigration records, Civil War pensions, homestead claims. All of that is real and valuable, but today we're staying closer to home. We're working in the 1900s and 1920s and 1930s, and I'm going to show you that those recent ancestors can be just as mysterious and just as hard to pin down as anyone who crossed the Atlantic 150 years ago. So let's get started. Every family has a version of this story. There's a grandmother or a great grandmother who existed in a kind of soft focus in the family's memory. You know the rough outlines. You know where she was from or where you think she was from. You know who she married or who you think she married. You know how many children she had or how many you've always been told she had. And for most families, nobody's ever questions it. Because why would you? The story was told by people who were there or at least closer to there than we are. It passed from mouth to mouth across the decades, and by the time it reaches us, it has the weight of fact. This is the case study I want to walk you through today. I'm going to call this woman Edna. Edna Louise Mercer, born sometime around eighteen eighty eight in Indiana. She married a man named Robert Harmon, they had children, they eventually made their way to Ohio, and Edna lived a long, quiet, unremarkable life by every outward measure. The family story, as it's been passed down, went something like this. Edna was an Ohio girl, born and raised there. She married Robert, had three children, kept a modest house, and was by all accounts a steady and private woman who did not talk about herself much. The children knew her as grandma, the grandchildren knew her as great grandma. She died before most of them could ask her the kinds of questions we now wish we had asked. Sound familiar? Here's what I actually knew going into this research session in terms of documented facts. I had a death certificate for Edna, dated nineteen sixty one, listing her birthplace as Indiana, not Ohio. That was the first crack in the story. Not Ohio, Indiana. Which could be a clerical error on a death certificate, and those happen all the time, so I filed it away. I had her in the nineteen twenty and nineteen thirty federal censuses living in Hamilton County, Ohio with Robert and what appeared to be three children. I had a marriage record for Robert and Edna from nineteen oh eight in Morgan County, Indiana. And I had a family tree that four generations of this family had contributed to, which showed exactly three children, one husband, and a birthplace of Ohio. That's what I thought I knew. Now, here's something I want to say before we get into the AI tools, and this is really the premise of today's entire episode. The family story isn't wrong because the people who told it were careless or dishonest. Family stories change because human memory simplifies. Human memory protects. Human memory across decades and generations tends to sand down the edges of painful things until they're smooth and manageable and easy to carry forward. What we're going to uncover today isn't a scandal. It's a sadness that got quietly set aside. And I want you to hold that frame as we go through this. AI is your research assistant, not your researcher. And what that means today is that the AI tools are going to help me see the evidence clearly. But the interpretation, the emotional understanding of what that evidence means for a real human life, that belongs to me. And eventually to the family. Okay, let me show you where this went. I went back to the nineteen ten census. I'd seen it before, but I hadn't looked at it carefully. I'd confirmed that Edna was there, confirmed she was married to Robert, and moved on. This time I looked at every column, and that's when I saw it. Children born five, children living four. The nineteen ten census had a column where enumerators were supposed to record whether someone was in their first or second marriage. When I looked at Edna's entry, the enumerator had actually filled it in. Marital status notation M two. Five children born, four still living at the time of the census, and a notation indicating this was not Edna's first marriage. I counted the children I could document. Three. I was missing at least one child who'd been alive in nineteen ten and at least one who had been born but was no longer living, and I had no record of any marriage before Robert. That's when I opened Claude. The first thing I want to say about why I started with Claude here and not one of the other tools is this. Before you can understand what's missing, you have to see exactly what you have. Claude is extraordinarily good at taking a set of documents or data points, organizing them into a structured comparison, and then identifying the gaps and inconsistencies with real analytical precision. That is exactly what I needed in this moment. I had data from three censuses, nineteen hundred, nineteen ten, and nineteen twenty. I had the nineteen oh eight marriage record, I had the death certificate, and I had the family tree information that had been passed down. What I didn't have was a clear side by side picture of how all of these records either confirmed or contradicted each other. So here's the exact prompt I typed into Claude. Prompt number one, the census comparison table. Quote I'm researching my ancestor Edna Luise Mercer, who later became Edna Harmon. I believe she was born approximately eighteen eighty eight in Indiana. Here's the data I have gathered from census records and other sources. nineteen hundred census listed as Edna Mercer, age twelve, birthplace Indiana, living with parents in Morgan County, Indiana. Parents listed as born in Indiana. nineteen oh eight marriage record, Morgan County, Indiana, Edna Mercer to Robert Harmon. nineteen ten census listed as Edna Harmon, age twenty two, birthplace Indiana, wife of Robert Harmon, age twenty six, Morgan County, Indiana. Marital status notation M two, children born five, children living four, three children are listed in the household ages zero, one, and three. nineteen twenty census, listed as Edna Harmon, age thirty two, birthplace Indiana, wife of Robert Harmon, Hamilton County, Ohio. Four children listed in household, ages eight, ten, eleven, and thirteen. nineteen thirty census, listed as Edna Harmon, age forty one, birthplace Indiana, Hamilton County, Ohio, two children remaining in household. Death certificate nineteen sixty one, lists birthplace as Indiana, lists survivor as husband Robert Harmon. Family oral history, born in Ohio, married Robert Harmon, had three children. Please create a detailed comparison table of every data point across all of these records. Then identify every inconsistency you can find, rank them from most to least genealogically significant, and for each inconsistency, give me at least two possible explanations I should investigate. Now, here is what Claude came back with. And I want to walk you through this carefully because this is exactly the kind of analytical output that turns a confusing pile of information into a genuine research roadmap. Claude built the comparison table. Across the top, each source, down the side every data point, name as recorded, age, birthplace, husband, number of children, marital history notation. Then it listed the inconsistencies ranked by significance. The most significant? The M2 marital status notation in 1910 combined with five children born and only three known children documented in the family tree. Claude flagged this immediately as the highest priority. A second marriage before 1908 means there was a first marriage, potentially with its own children, of which only partial evidence survives in the census count. Second most significant, the nineteen twenty census shows four children in the household ages eight, ten, eleven, and thirteen. If Edna married Robert in nineteen oh eight, the thirteen year old would have been born in approximately nineteen oh seven before the documented. That child is either from the first marriage, a stepchild of Roberts, or the nineteen oh eight marriage date is not actually the beginning of their relationship. Third, the Born in Ohio family story contradicts every primary source which consistently lists Indiana. And then Claude offered its explanations. For the first marriage, two possibilities. Edna was widowed young and the first husband is simply absent from family memory. Or the first marriage ended in divorce, which carried social stigma in rural Indiana in the early 1900s, which could explain why it was never discussed. For the Ohio birthplace story, Claude suggested the family may have conflated where Edna lived in later life with where she was born, which is an extremely common intergenerational distortion of oral history. Or the family may have been told she was from Ohio by Edna herself as a form of privacy around her earlier life. That last line from Claude hit me in a way I didn't expect. A form of privacy around her earlier life. Because now I wasn't just looking at inconsistent records. I was looking at a woman who may have chosen deliberately and quietly to close a chapter. And I needed to find out what was in it. My first instinct when I saw the Ohio discrepancy was to search Ohio, which is exactly the kind of logical but wrong move that genealogy research tempts you into constantly. The family says Ohio. The records say Indiana. Maybe there was a brief period as a child when the family lived in Ohio. Maybe the nineteen hundred census has an heir. So I went looking. I went to the Federal Census Records. I checked the nineteen hundred Ohio Federal Census, searching for any Mercer family with a daughter of the right age. I checked the eighteen eighty Ohio Federal Census for Edna's parents. I searched Ohio City Directories from the eighteen nineties that are freely available on archive.org. Nothing. No Mercer family in Ohio that matched Edna's profile in any of those sources. I had spent forty minutes chasing Ohio and come up completely empty. And that is when I had to stop and ask the harder question. Not where was the evidence wrong, but where was my assumption wrong? The evidence consistently said Indiana. Every primary source that would have captured her birthplace from someone who knew her, including her own census entries where she presumably provided her own information, said Indiana. The only thing saying Ohio was the family oral history. And as I already started to understand, that oral history has other holes in it. So I set Ohio aside and I did something I should have done 20 minutes earlier. I asked for help thinking through this. This is where ChatGPT came in, and I want to explain why I switched tools here. Cloud is exceptional for document analysis, for comparison, for pattern recognition across a specific set of data. But when I need to brainstorm, when I need to generate a list of possibilities I haven't thought of, when I need creative lateral thinking about research directions, that is where ChatGPT earns its place in the toolkit. These tools have different strengths, and the skill is knowing which strength fits which moment. Here's what I asked. Prompt number two, the brainstorming research checklist. Quote I'm researching a woman in rural Indiana in the early 1900s. Her 1910 census entry shows she was in her second marriage, marked M2, and had five children born with four still living. My documented tree only shows three children from her second marriage, which began in 1908. The family oral history never mentioned a first marriage or any additional children. The family also consistently described her as being from Ohio, but every primary document lists Indiana as her birthplace. Please answer these questions. two, what specific records should I search to find evidence of that first marriage and where can I access them ideally for free? three, the nineteen twenty census shows a child in the household who would have been born before the documented nineteen oh eight marriage. What does this suggest and what records would clarify this? four, why might a family consistently misremember a birthplace across multiple generations, substituting a later place of residence for a place of birth? Please give me a prioritized research checklist, end quote. ChatGPT's response was thorough and genuinely useful. Let me walk you through the most important parts because this is the kind of output that reframes your entire research approach. On why a first marriage would disappear from family memory. ChatGPT outlined four common historical patterns. Early widowhood, especially from industrial accidents, illness, or the complications of early childbirth. Was common enough in the 1900s and 1910s that remarriage within a year or two was not unusual. The surviving spouse often simply moved forward and the first marriage faded. Second? If there were children from the first marriage who went to live with the first husband's family or with other relatives, those children were literally absent from the daily household. Making it easier for the second family to form around the new marriage without constantly referencing what came before. Third, social attitudes Third, social attitudes towards divorce in rural Indiana in their early 1900s were such that a divorced woman faced genuine community judgment and protecting her reputation and her children's futures gave real motivation to keep the history quiet. Fourth, Edna herself may have chosen not to discuss it. And one generation's silence mean and one generation silence means the next generation's ignorance. On research priorities. Family search first, specifically the Indiana Marriages Collection covering 1811 through 2019, which is free and searchable by name. Search for Edna Mercer as the bride with a date range of approximately 1903 to 1907, before the 1908 Robert Hammond, before the 1908 Robert Harmon marriage, county level marriage records in Morgan County, Indiana, and neighboring counties, then Indiana death records from 1900 to 1910 for any male with the right surname who might be the first husband. Then the 1910 census, specifically looking for any child in the household or in a neighboring household with a different surname that might represent a child from the first marriage. On the 1920 childborn before 1908, ChatGPT flagged this as strongly suggesting the child from the first marriage was eventually absorbed into the Robert Harmon household, either formally through adoption or informally, and that searching for that child's birth record under the first husband's surname was the next logical step. On the birthplace confusion, this is, according to ChatGPT, one of the most consistent patterns in intergenerational oral history transmission. A family moves from Indiana to Ohio. The children grow up in Ohio. Grandma lives in Ohio for the last thirty or forty years of her life. By the time the grandchildren and great grandchildren tell her story, Ohio is where she lived for the vast majority of their experience of her, and Indiana has become an abstraction. The birthplace and the life place collapse into one in the family memory. It's not dishonesty, it's how memory works across time. I want to pause here for a moment because this is actually a GPS point worth naming explicitly. What I was doing right now, working through these ChatGPT results and thinking about which research paths to pursue is GPS element one. Reasonably exhaustive research doesn't just mean searching every database. It means thinking systematically about which records could exist and why, and then searching those records deliberately rather than at random. AI helps you build that research map before you start driving. Now I had a research checklist, and I went to the first item on it. Family search. Indiana Marriages. Edna Mercer as Bride. Date range 1903 to 1907. And this is where I switched to perplexity, and I want to tell you exactly why. Perplexity is built for cited research. When I need to understand what records exist in a specific Specific state for a specific time period, and I need source verified information rather than assumptions. Perplexity is the tool I reach for. I didn't need it to search for Edna specifically. I needed it to help me understand the landscape of what Indiana records actually survive from this era and what I could realistically expect to find. Here's what I searched. Prompt number three Verifying record availability. Quote What Indiana genealogy records from approximately nineteen hundred to nineteen fifteen are available for free online? Specifically, marriage records, death records, and birth records. Which counties are well covered versus poorly covered? What is available on family search versus ancestry versus state archives? What were Indiana's requirements for recording vital events during this period in quote? Perplexity confirmed what I had found in my own preliminary research and added precision. Indiana did not require birth and death records to be kept at the state level until nineteen oh seven. Before that, county health offices were supposed to maintain them, but compliance was inconsistent. Marriage records, however, were kept at the county clerk level going back to statehood, and Family Search has digitized Indiana marriages from 1811 through 2019, with coverage that varies by county, but it's generally solid for Morgan County in this period. Armed with that confirmation, I went directly to Family Search. Indiana marriages. I searched for Edna Mercer as bride with a marriage year range of 1903 to 1907 in Indiana. And there she was. And I want to say here, Morgan County's marriage records from this period are among the better digitized in Indiana. Not every county would give you this result. I got fortunate, but I found her. Edna L. Mercer, Bride. Aldrich, Morgan County, Indiana, nineteen oh five. Her first husband's name was Floyd Aldrich. They married in nineteen oh five. She was approximately seventeen years old. I sat with that for a minute. Seventeen. Married in nineteen oh five, and then married again to Robert Harmon in nineteen oh eight, three years later. Something happened to Floyd. I want to tell you something about Indiana in nineteen oh five before we go any further, because the historical context here matters. Morgan County in the early nineteen hundreds was a rural agricultural county about thirty miles southwest of Indianapolis. Life expectancy for young men doing farm labor in that area was significantly shorter than it is today. Typhoid fever was a genuine and recurring killer in rural communities throughout Indiana well into the second decade of the twentieth century. It spread through contaminated water supplies, through food, through contact with infected individuals. A young man working on a farm or near a poorly maintained well in 1906 or 1907 was genuinely at risk in a way that is hard for us to conceptualize in 2026. I went looking for Floyd Aldrich in Indiana Death Records. The Indiana State Archives website notes that death records before 1907 are inconsistently preserved at the county level. But the 1907 and later records, which the state required to be filed, are more accessible. I found Floyd H. Aldrich in the Indiana Death Index Morgan County 1907. Cause of death listed as typhoid fever. Age at death twenty two. Floyd Aldrich died of typhoid fever in nineteen oh seven at age twenty two, leaving Edna a widow at approximately nineteen years old. I want you to hold that image for a moment. Nineteen years old. Her husband dead of fever, at least one baby either already born or coming. In a rural Indiana county in nineteen oh seven, with no social safety net, no widow's benefits, and a family whose options were limited by the economics of farm life. And now the M two notation made complete sense. And now the five children made sense, though I still didn't know where all of them were. And now the silence made sense. Edna Mercer had married Floyd Aldrich at seventeen, became a widow at nineteen, and remarried Robert Harmon at approximately twenty. Somewhere in that two year window between Floyd's death and the Harmon marriage, at least one child had been born. Possibly two. She had done what she had to do, and then she had moved forward, and moving forward had eventually involved not looking back. And the family had never said a word about any of it. This is the moment in genealogy research that I find the most humbling, and I think a lot of you know exactly what I mean. You start looking for a date or a place. You end up looking at a life. The records don't only give you information, they give you a window into a set of decisions that were made under pressures and circumstances that we can barely imagine from where we stand today. That is why we do this work. That is why it matters. This is where Notebook LM became essential, and I want to take a moment to explain what Notebook LM does that none of the other tools in this episode can do because it's a genuinely different kind of tool. Notebook LM is not a chatbot. It doesn't draw on the broader internet or on its own training knowledge to answer you. It works exclusively from the documents and sources you upload to it. That is its entire design philosophy. And for genealogy, that is extraordinarily powerful because it means when you ask Notebook Elm a question, every answer it gives you is anchored to your actual evidence. It cannot hallucinate a fact that isn't in your documents because it is only working from your documents. What I did was create a notebook Elm notebook and upload everything I had gathered. The 1900 census image for Edna, the 1910 census image, the nineteen twenty and thirty census images, the nineteen oh five marriage record for Edna Mercer and Floyd Aldrich, Floyd's death record from nineteen oh seven, the nineteen oh eight marriage record for Edna and Robert Harmon, and a plain text document where I had typed out the family oral history as I understood it. Then I asked Nobook LM to do something very specific. Prompt number four, the evidence timeline and gap analysis. Quote Based only on the records I have uploaded, please build a complete chronological timeline of Edna's life from the earliest record to the latest. Then identify every gap in the timeline where a significant life event, such as a birth, marriage, or death of a family member, might have occurred but is not documented in these sources. For each gap, tell me which specific record type would most likely fill it and where I could find that record in quote. Remember, AI is your research assistant, not your researcher. And that is never more true than with Notebook LM. Because what it returned was a synthesis of my evidence, and where it identified gaps, those were genuine gaps in my evidence, not gaps it had invented. The timeline it built was clean and precise. eighteen eighty eight approximate birth, Morgan County, Indiana. nineteen hundred living with parents in Morgan County. nineteen oh five married Floyd Aldrich. nineteen oh seven, Floyd Aldrich died. nineteen oh eight married Robert Harmon. nineteen oh nine through nineteen eleven, three children born based on their ages in the nineteen ten census. And then the gap that Nobook Ellen flagged as the most significant. Between nineteen oh five and nineteen oh eight, the timeline was nearly empty. The nineteen ten census showed five children born to Edna. The three youngest were evidently from the Harmon marriage. The gap between nineteen oh five and nineteen oh eight, the Aldridge years, was almost entirely undocumented in my uploaded sources. Nobook LM identified that at least one and possibly two children would have been born during this period based on the census arithmetic, and that I had no records for those children at all. It suggested Indiana birth records for Morgan County nineteen oh five to nineteen oh eight. And then it connected something I hadn't consciously put together yet. The nineteen oh five Floyd Aldrich marriage record listed Morgan County as his residence. Nobook LM linked his surname to the undocumented children gap and suggested I searched the nineteen ten census specifically for any households with the surname Aldrich in the same county where Floyd had lived. I went back to the nineteen ten census, not to find Edna, but to look at the household near Edna's parents in Morgan County, and that is where I found her. An entry in the household of one Frank Mercer, who I identified from the nineteen hundred census as Edna's father, living with Frank and his wife was a child listed as Nellie Aldrich, H four, born approximately nineteen oh six, surname Aldrich. Nellie Aldrich, four years old, living with Edna's parents, surname of Floyd's family. Edna had left her daughter with her own parents when she remarried Robert Harmon. Now I want to be clear about something here because this is where responsible genealogical interpretation matters. I am not saying Edna abandoned her daughter. The records don't support that conclusion, and more importantly, neither does the historical context. Leaving a child with grandparents while a young widow rebuilt her life was not uncommon in rural America in nineteen oh eight. It was sometimes the most loving thing a mother in that situation could do. Edna's parents were established in Morgan County. They had a home, a farm, stability. A nineteen year old widow starting over in a new marriage had none of those things yet. The decision, if we can even call it a decision in the way we'd use that word today, was probably made out of grief and necessity in equal measure. I sat with that for a long time. Sat with what that must have meant a nineteen year old widow, a baby girl, a chance to start over, and a decision made in grief and survival to give her daughter to her parents to raise while she built a new life with a new husband. That's not a scandal. That's not a secret kept out of shame. That's a human being doing what she had to do to survive a situation that had cost her everything, and then carrying the loss quietly for the rest of her life. Ohio wasn't a lie. Ohio was where she eventually built the life she was trying to build. The three children she raised with Robert were real. But so was Nelly. And so was Floyd. And so was the girl who became a widow at nineteen in Morgan County, Indiana and started over. The unexpected discovery here isn't a dramatic revelation. It's something quieter and more honest than that. The family story wasn't wrong because anyone tried to deceive the next generation. It was incomplete because Edna herself may have chosen not to reopen a wound that had taken years to close. Her silence became her children's ignorance, and her children's ignorance became the family story that reached us. Now we know. Now here is your actual homework this week. Go find one family story, just one, that you've accepted at face value. Maybe it's a birthplace. Maybe it's a marriage count. Maybe it's a number of children. Maybe it's a date that the family has always stated confidently. Pick one claim from your oral history and run it against every primary source you can find for that ancestor. Start with the census records. If your ancestor was a woman living between 1900 and 1920, pull both the 1900 and 1910 censuses if she appears in them and look at those two numbers. Children born, children living. If the numbers in those two columns don't match exactly what your family tree shows, you have a thread worth pulling. If there's an M two notation and you only knew about one marriage, you have a thread worth pulling. These clues are sitting there in plain sight on records that are freely available on Family Search right now. Then bring that discrepancy to Claude and run the comparison table prompt from today's episode. That prompt alone, the one asking Claude to compare every data point across multiple records and rank the inconsistencies by genealogical significance, is something you can use on virtually any ancestor in any era. It doesn't require a mystery like Edna's. It works just as well when you're trying to confirm something you thought you already knew. Not to prove the story wrong, to see what the evidence actually says. Sometimes the story is right. Sometimes it's close but garbled. And sometimes you find a little girl named Nelly living with her grandparents, and you understand for the first time what your ancestor carried for 60 years without ever saying a word. That's the work, and it's worth doing. And for my Australian and UK listeners, every technique we cover today translates directly to your research. If you're working with Australian ancestors from the early 1900s, census substitutes like electoral rolls, which are free on the National Archives of Australia's website at naa.gov.au can do much of what the US Census does in these comparisons. State births, deaths, and marriages registries hold marriage records that can surface a first marriage the family never mentioned. For UK researchers, the England and Wales Civil Registration System has indexed marriages from 1837 onward, and free BMD at freebmd.org.uk gives you free access to birth, marriage, and death indexes that can reveal exactly the kind of discrepancy we found today. The story that Grandma never told exists in British and Australian families just as it does in American ones. The records no. Thank you so much for listening to Ancestors and Algorithms. If today's episode resonated with you, and I have a feeling it did, please leave a review wherever you listen to podcasts. And if you know a fellow genealogist or family history researcher who has ever bumped up against a family story that didn't quite add up, share this episode with them. That is the best way to help our community grow. For my Patreon members, the companion guide for this episode is now in your library. You'll find 12 advanced prompts that take today's techniques to the next level, including a multi-step workflow for using Claude and Nobook LM together to build a complete evidence conflict report on any ancestor, a prompt for researching remarriage and blended family patterns in early 20th century records, and a full GPS research checklist tailored specifically to the family story verification process. And if you've been thinking about joining, I host a monthly live QA sessions on YouTube exclusively for members, where you can bring exactly the kind of question today's episode raised. Head to ancestorsnai.com to learn more. 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.