Ep. 41: Tracing Enslaved Ancestors with AI | Using Perplexity, Claude, NotebookLM, and ChatGPT to Navigate the 1870 Wall

Brian uses 4 AI tools to trace a formerly enslaved Mississippi ancestor across 6 record collections and through the 1870 Wall. This is what African American genealogy research looks like when AI and the Genealogical Proof Standard work together.

If your ancestor was enslaved, the census did not record their name before 1870. Every year before that belongs to a completely different research strategy, and today you will learn exactly what that strategy looks like in action.

What you will learn in this episode:

• How to use Perplexity to map the specific records that survived in your ancestor's county before you search a single database
• How to use Claude to analyze Freedmen's Bureau documents, labor contracts, and estate inventories for clues you would otherwise miss
• How to upload multiple documents to NotebookLM and build a source-grounded evidentiary timeline that only draws on what you can actually prove
• How to use ChatGPT to brainstorm every explanation for why an ancestor disappears from the record after 1880
• How to use the 1860 slave schedule and probate records to connect a formerly enslaved ancestor to a specific property before emancipation
• What the FAN club method (Family, Associates, and Neighbors) looks like in this era, and why it is the primary tool for breaking through the 1870 Wall
• Why the Freedman's Bank records on FamilySearch are free to search and can contain more personal detail than a dozen census entries combined
• What honest research in this area looks like, including what AI cannot do, and what the silence in these records actually means

This episode ends as a partial answer. The wall is thinner. It is not gone. That honesty is the point.

For Australian and UK researchers: these techniques apply directly to Aboriginal and Torres Strait Islander ancestry research through AIATSIS and state records offices, and to British colonial slavery research through the Legacies of British Slavery database at UCL and The National Archives at Kew.

The Companion Guide includes 12 advanced prompts for African American genealogy research, a multi-step 1870 Wall workflow, a GPS Research Checklist, and a guide to every Freedmen's Bureau record type. Available to Patreon members at ancestorsandai.com.

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 page in a Freedman's Bureau Register from Notches, Mississippi, dated september eighteen sixty five. Halfway down in faded ink is a name Hezekiah Coles, age listed as twenty seven, former occupation, field labor, and then in the column marked former owner, a name that is not his own. That one line is everything I had to start with. No birthplace, no parents, no wife or children listed, a name, an age, and the ghost of someone else's property record standing behind him. What I found next took me through six different record collections, four AI tools working in tandem, and one assumption that almost cost me the whole case. Let's dive in. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host Brian and I want to start today a little differently. I've been getting messages from some of you asking me to cover African American genealogy research, how to use AI tools when your ancestors were enslaved, how to navigate the records that exist and make peace with the records that don't. I'm grateful you asked, and I want to be up front about something before we start. I'm a white man, I grew up with family trees that were inconvenient sometimes, complicated sometimes, but never deliberately destroyed. My ancestors were recorded. Hezekiah Cole's was not. So I want to approach today with the humility that requires. I am here to share tools and methods, not to speak with authority about what this search means to the people doing it. What I can tell you is that researching across the eighteen seventy wall, which is what genealogists call the point before which most enslaved people have no official record of their existence, is among the most difficult and also the most important work in family history. The wall was put there on purpose by a legal system that treated human beings as property and had no interest in recording their names, their families, or their lives. That context is not a prelude to the research. It is the research. You can't understand these records without understanding what they refuse to preserve. Today we're going to work through that wall together, honestly, including where the tools fall short and where the records end before we are ready for them to. AI can help with this research in meaningful ways, but it can't give back what was taken. So let's get started. I want to tell you about a man named Hezekiah Coles. I found Hezekiah in the eighteen seventy Federal Census living in Adams County, Mississippi, in the area around Notches. He was listed as thirty two years old, a farm laborer, born in Mississippi. He was living in a household with a woman named Temperance, listed as twenty eight and three children, Moses seven, Lula four, and an infant named Aaron, all born in Mississippi. Now, if you know anything about African American genealogy, you already know what the eighteen seventy census means. It was the first federal census to list all formerly enslaved people by name. Before eighteen seventy, if your ancestor was enslaved, the census didn't record their names at all. The eighteen sixty and eighteen fifty slave schedules listed enslaved people by age, sex, and color only, under the name of whoever held them in bondage. No names, no families, just numbers in a column. I want to sit with that for a moment because I think it can slide past us when we say it quickly. The United States government kept detailed records of the ages and physical descriptions of millions of human beings and categorized all of it under someone else's name. Hezekiah Coles existed for 27 years before 1865, and the federal government's official position was that he didn't need a name recorded. His children's birth years, his parents, where he grew up, all of it was left to silence. So the 1870 census is often the first official record a formerly enslaved person ever had that acknowledged their existence as an individual human being with a name. That is not a small thing. That is everything. For those of you who are researching your own family through this wall, I want to acknowledge something. Finding a great great grandparent in the eighteen seventy census for the first time, knowing they were thirty two years old before anyone wrote their name in a government record is a different kind of moment than most genealogy discoveries. I can't fully understand what it means to find that page, but I understand it deserves more than a checkbox in a research log. And it also means that for Hezekiah Coles, I was starting with a record from when he was thirty two years old. Every year before that, every record that might tell you who his parents were, where he was born, whether he had siblings, who raised him, what his life looked like before emancipation, all of that sits behind what genealogists call the eighteen seventy wall. The point in time before which the paper trail for formerly enslaved people essentially disappears. Getting past it requires a completely different research strategy. You cannot work backward census by census the way you can for most ancestors. You have to reconstruct the world around him, find the people who were connected to him, identify which slaveholders' records might contain some trace of his existence, and then comb through those records looking for any evidence that he was there. It is painstaking work. It requires patience I'm still developing. This is where AI changes the game. Not because AI can find records that don't exist, and I want to be crystal clear about that, but because AI can help you process enormous amounts of historical context, identify research pathways you might not have considered, and analyze fragmentary evidence with a precision and speed that would take a human researcher months to replicate on their own. Remember, AI is your research assistant, not your researcher. And no type of genealogy is that more important to hold on to than this one. Here's what I knew going in. I had Hezekiah in the eighteen seventy census, Adams County, Mississippi. I had a reference to him in a Freedman Bureau Register from 1865, which gave me his approximate birth year as eighteen thirty eight, his former occupation, and a former enslaver surname. I also had a labor contract from late eighteen sixty five, which confirmed he was working in the Nachez area after emancipation. What I didn't have his parents, his wife Timpress' maiden name, any record of where he was enslaved specifically, and most significantly I had no record of what happened to him after eighteen eighty when he stops appearing. That gap after eighteen eighty was almost as troubling as the gap before eighteen seventy, and it would turn out to be the more complicated mystery. Let me walk you through exactly how I tackled this. My first move was perplexity, and I want to explain exactly why I started there rather than jumping straight into the records. When you're doing African American genealogy research in the Deep South in this era, you need to understand the specific record landscape of that county at that time. Not Mississippi in general. Adams County specifically. What offices functioned? What burned? What survived? What has been digitized versus what requires a trip to a physical archive? That context shapes your entire strategy and getting it wrong waste weeks. So I opened perplexity and here's the exact prompt I used. Quote I am researching an African American man named Hezekiah Coles who appears in Adams County, Mississippi Records between eighteen sixty five and eighteen eighty. He was formerly enslaved and appears in the eighteen seventy Federal Census in the Natchez area. What specific genealogical record collections survive for Adams County, Mississippi that would be most useful for pre-1870 African American research? Include Freedmen's Bureau Records, Freedmen's Bank Records, Slave Schedules, Plantation Records, Church Records, and any digitized collections available through Family Search, Ancestry, or the Mississippi Department of Archives and History. Also explain which records are or not available online as of 2026. What perplexity returned was a detailed cited breakdown that saved me probably three hours of independent digging. Let me share the highlights because the context is directly relevant to your own research if you have Mississippi ancestors. The Mississippi Freedmen's Bureau filled office records from 1865 to 1872 are available through Family Search as images, but here's the important caveat. They are not fully indexed. Many of the roles contain names that are visible in the images but have not been pulled into a searchable database. That means you may need to browse the images directly, not search by name alone. The Freedmen's Bank had a notches branch. The bank records from 1865 to 1874 are indexed on Family Search and they are genuinely remarkable. When a depositor opened an account, the bank recorded their age, complexion, occupation, and birthplace, and often the names of their parents, siblings, and children. Sometimes, and this is the detail that can break a case wide open, they also recorded the name of their former enslaver. One bank record can do the work of a dozen census entries. Perplexity also flags something I hadn't considered. The 1860 slave schedules for Adams County are available on Family Search and Ancestry. They do not name enslaved individuals, but they do list the enslavers' name, the county, and the age, sex, and complexion of each enslaved person. If I knew the enslaver's name from Hezekiah's Freedman's Bureau Record, I could cross-reference the 1860 slave schedule to see whether a male matching Hezekiah's approximate age in 1860 appears in that enslaver's household. That was the play. And it required me to go back to the bureau record and look more carefully at that name in the former owner column. So I did. And what I found required me to completely rethink my approach. The name in Hezekiah Friedman's Bureau Register was listed as T Mercer. That's it. One initial, one surname. In eighteen sixty five, Notches was a major city with a significant population. There were almost certainly multiple people with the surname Mercer in Adams County and the surrounding counties. One initial was not going to get me far on its own. I needed to figure out which T Mercer was most likely to have been in a position to hold people in bondage in this part of Mississippi before the Civil War. This is where I brought in Claude. I uploaded three documents, the Freedmen's Bureau Register image, the labor contract from late 1865, and a typed transcript I had made of what I could read in both. Then I gave Claude this prompt. Quote, I am researching an African American man named Hezekiah Coles who was formerly enslaved in Adams County, Mississippi. His Freedmen's Bureau Register from 1865 lists his former enslaver as T Mercer. The attached labor contract from late 1865 shows he agreed to work as a farm laborer on a property in Adams County. I need help developing a research strategy to identify which T Mercer in Adams County, Mississippi most likely held him in bondage before emancipation. Please analyze the documents for any additional clues I may have missed, suggest what records I should search to narrow down which T Mercer this was, and explain the reasoning behind each suggestion. Note I have access to Family Search, Ancestry, and the Mississippi Department of Archives and History online collections end quote. Claude's response was methodical in exactly the way I needed. It pulled out two details from the documents I had glossed over. First, the labor contract included a brief property description that mentioned proximity to the Mississippi River, which helped narrow the likely location within the county. Second, the labor contract had a witness signature, which Claude flagged as potentially significant because witnesses to labor contracts in this period were often neighbors or local officials who might appear in the same geographic cluster of records. Claude then laid out a research pathway. Search the 1860 Adams County slave schedules for all Mercer surnamed slaveholders. Cross-reference any Mercer entries against the 1860 Free Population Census to establish first names, ages, and occupations. Check probate records for the Mercer family because estate inventories from this period sometimes named enslaved individuals. And check the labor contracts filed with the Mississippi Freedmen's Bureau Assistant Commissioner in 1865 and 1866, which sometimes list full names and ages of all parties and could confirm which Mercer was Hezekiah's former enslaver. This is exactly what I mean by GPS in action. That is GPS element three, the correlation of evidence across multiple source types happening in real time as Claude mapped connections between fragmentary documents. I followed the first step. In the 1860 slave schedule for Adams County, I found two Mercer households. One, a Thomas Mercer, listed eleven enslaved individuals. Among them was one male listed as approximately twenty two years old. If Hezekiah was born around 1838, as his bureau record suggested, he would have been approximately twenty two in eighteen sixty. The ages matched. The other Mercer in the county held only two enslaved people, both female. That narrowed it considerably. Thomas Mercer was almost certainly the T Mercer in Hezekiah's record. Now I want to tell you where I hit a genuine dead end. Because this is important. I went looking for the Freedman's Bank records for nachos. The branch existed. The records are on family search. I searched for Hezekiah Coles and came up with nothing. I searched variations on the name, Hezekiah, H. Coles, Coles with different first names. Nothing from Adams County that matched. Not every formerly enslaved person opened a bank account. The bank collapsed in 1874 and many depositories lost what little they had saved. The absence of a record is not evidence of absence from the county. It is a closed door, nothing more. So I regrouped. Claude had mentioned probate records at the next step. I went looking for Thomas Mercer's estate records in Adams County. Here's where the research took an unexpected turn. Thomas Mercer died in 1869. His estate went through probate in Adams County in 1870 and 1871. Those records are available through family search for Adams County. And in the estate inventory, among the itemized debts and assets, there is a note about money owed to a freedman listed as Hezekiah for labor performed on the property after emancipation. He was still there. Four years after walking out of bondage, Hezekiah Coles was still working the same land. I've thought about that a lot since I found this record. The reasons that kept formerly enslaved people tied to the same ground after emancipation were not simple. Some had nowhere else to go. Some had families still on the property. Some had no money, no transportation, and no knowledge of what waited for them anywhere else. Freedom in eighteen sixty five did not come with a road, a house, or a bank account. For many people, it came with a labor contract and the same field they had worked a year before, except now theoretically for wages. I say theoretically because the wages were often withheld, contested, or never fully paid. The estate inventory records money owed to Hezekiah, not money paid, owed. I'm not telling you this to make the episode harder than it needs to be. I'm telling you because if you are researching your own family through this era, you may find your ancestor in the same situation. And understanding why they stayed on the same land is part of understanding who they were and what kind of strength it took to build a family, find a wife, name children, and keep going in those years. Hezekiah and Temperance had a four-year-old daughter named Lula and an infant son named Aaron by 1870. They were building something. That matters as much as the records. The estate inventory confirmed his connection to the Mercer property and gave me the documentary bridge between the pre-1870 slave schedule and the post-1870 census. I now had evidence placing Hezekiah Coles in a specific geographic and social context across multiple record types. That is what the GPS calls a reasonably exhaustive correlation of sources. And AI helped me map the pathway to get there. This is where I brought in Notebook LM. I had accumulated six separate documents by this point. The Bureau Register, the labor contract, the 1860 slave schedule entry, Claude's analysis summary, the estate inventory reference, and the 1870 census entry. I uploaded all of them to a single Notebook LM notebook and asked it to help me build a coherent timeline. Quote: Based only on the six documents I've uploaded, create a chronological timeline of what can be documented about Hezekiah Cole between 1838 and 1870. For each entry in the timeline, identify which specific document supports it and note whether the evidence is direct, indirect, or circumstantial. Also list any contradictions or gaps between the documents that I should address. I want to stop here and explain why I use Nobook LM for this rather than Claude, even though Claude is my go-to document analysis tool. This is the tool to function lock in practice. Notebook LM's fundamental design principle is that it answers only from the documents you provide. It doesn't reach outside those documents. It does not speculate. It does not add background knowledge. When I'm building an evidentiary timeline and I need to know that every claim is grounded strictly in the documents I have gathered rather than AI generated background information, NoBook LM is the right tool. It forces the discipline of source grounded thinking, and it catches gaps that feel filled in but are not actually documented. The timeline Nobook LM produced was clean, honest, and useful, and it flagged one gap I'd been mentally glossing over. I had no documented information about Temperance Coles. Not her maiden name, not her age before eighteen seventy, nothing that connected her to Hezekiah before the eighteen seventy census listed them in the same household. That absence is significant. If Temperance had a different former enslaver than Hezekiah, she might appear in completely different record collections. Her family connections might be the key to understanding what happened to this family after 1880. And that is when I realized I'd been chasing the wrong thread. I had focused entirely on Hezekiah's pre eighteen seventy story, but the deep The super mystery was the post 1880 disappearance. After I found him solidly in 1870, I expected to find him in 1880. But the 1880 census for Adams County doesn't show a Hezekiah Coles matching his age. There is a Coles family in the county, but the ages don't line up. I took this to ChatGPT, not for document analysis, but for what it does best, brainstorming alternative explanations and research pathways when you're stuck. Quote, I'm researching Hezekiah Coles, a black man born approximately 1838 in Mississippi, documented in Adams County in 1865 and 1870, but who I cannot locate in the 1880 census. His wife was Timperance, approximately born 1842. They had at least three children in 1870, Moses seven, Lula four, and Aaron Infant. What are the most common explanations for why a black man in this time period and region might not appear in the eighteen eighty census despite surviving past eighteen seventy? Please list research strategies I could use to locate him or establish what happened, including migration patterns out of Mississippi during Reconstruction, name changes, enumeration errors, and any other possibilities relevant to the specific time and place end quote. ChatGPT gave me eight possibilities. Death between 1870 and 1880 was one. Migration was a real possibility. Many black families left Mississippi for other southern states or for the Midwest during and after Reconstruction, which came to a head in Mississippi in 1875 when the Mississippi Plan systematically drove black voters from the polls and stripped away the political gains of the preceding decade. Enumerator errors were common, with names misspelled or recorded under initials only. Some men appeared in a household as a laborer without being identified as head of household, meaning their first name might not be captured. And then there was one I had not considered at all. What if Hezekiah and Temperates had separated? If he left Adams County and she stayed, they would appear in different census records, and I might have been searching for a household that no longer existed as a unit. ChetupT suggested I searched the eighteen eighty census not for Hezekiah Coles specifically, but for Temperitz Coles and any of the children by name. The children, particularly Moses, born around eighteen sixty three, would be approximately seventeen and eighteen eighty, old enough to appear in records independently. I searched. And I found Temperance Coles, listed as a widow in Adams County in the eighteen eighty census. With two children, Lula now fourteen and Aaron ten. Moses was not in the household. Hezekiah Coles appears to have died between 1870 and 1880. I have not found a death record. Adams County records from this period are incomplete. But Temperance's status as a widow in 1880 is direct evidence that he was gone. I know that's not the resolution I was hoping for. And it is not the resolution you were probably hoping for either. I'm going to be honest with you. I didn't find a triumphant breakthrough. I found the edges of what the surviving records can tell us. But here's what I want you to hold on to. So here's where we stand. We now know that Hezekiah Coles was born approximately eighteen thirty eight, almost certainly in Mississippi, and was enslaved in Adams County on the property of Thomas Mercer. We know he survived the Civil War. We know he registered with the Freedmen's Bureau in september eighteen sixty five and signed a labor contract that year. We know that by eighteen seventy he had a wife named Timperance and at least three children. We know he appeared in the estate records of his former enslaver working the same land for wages he may never have fully received. And we know that he died before eighteen eighty. Timperance was listed as a widow. She had two children with her, Lula fourteen and Aaron Tin. Moses the oldest was gone. We don't know where we don't know why. We only know he was not there. Timperance Coles raised two children alone in Adams County, Mississippi, in the years when Reconstruction was being dismantled around her. No husband, one child unaccounted for. The records do not tell us what that cost her. They only tell us she was there. I find myself genuinely moved by the fragments we have. Hezekiah chose a name for himself or kept the name given to him that carried the weight of the Old Testament. Hezekiah was a king who reformed his kingdom and trusted that something greater than himself would carry him through. Whether that name was chosen with that meaning in mind, I cannot know. But I notice it. He named his sons Moses and Aaron. He was building a family with names that held history. That was its own kind of declaration. What we do not know still outnumbers what we do. His parents' names, his birth country or precise location, where Temperance came from. What became of Moses? What Hezekiah thought and felt and hoped for in the years between eighteen sixty five and his death. The records cannot give us that. And I want to be honest, no AI tool, no matter how sophisticated, will ever recover what was deliberately taken. The silence in these records is not a gap to be filled. It is the evidence of what was done. What we can do, and what I believe we owe to the people we are researching, is search as thoroughly as the surviving record allows. By establishing the Mercer connection, we opened a door. The other people enslaved on that property, if any of them survived and left their own paper trail, may be connected to Hezekiah. That is the fan club approach, the family associates and neighbors cluster, and it is the methodology that gives this research its best chance. You research everyone in that social orbit, not the ancestor alone. AI tools make that kind of wide knit cluster research feasible for one person working alone in a way that was not possible before. Remember the golden rule AI is your research assistant, not your researcher. In this kind of research, that distinction matters more than anywhere else. AI can't see through what was deliberately erased, but it can help you navigate what survived with a care and a precision that honors the effort this search deserves. Hezekiah Coles lived. He had a name, a family, a faith strong enough to name a son Moses and another Aaron. He worked land that had held him and kept working it after it could no longer own him. He deserves the best research we can bring to the table. And so does temperance. And so does Moses wherever he went. Your homework, if you have African American ancestors who appear for the first time in the eighteen seventy census, start building a research file on them right now. Use perplexity to establish what records survive in their county. Search the Freedmen's bank records on Family Search, they are free to access and sometimes contain more personal detail than any other record from this period. And if you can identify a probable former enslaver, begin building a research file on that family's records. The cluster is where the answers live. And for my Australian and UK listeners, the techniques we cover today translate directly to research on marginalized communities in your archives. Australian researchers dealing with Aboriginal and Torres Strait Islander ancestry should start with IATIS, which is the Australian Institute of Aboriginal and Torres Strait Islander Studies at AIATSIS.gov.au, a sensitive but important resource for this research. The State Records Office, particularly in New South Wales and Queensland, also hold records on Aboriginal missions and stations that function as a parallel to the Freedmen's Bureau Records. UK researchers working on records of enslaved people connected to British colonial plantations should explore the legacies of British slavery database at UCL.ac.uk forward slash LBS, which documents enslavers and their estates with significant genealogical detail. The National Archives at Q also holds a colonial office records that sometimes include named individuals. Same principles, different archives. Thank you so much for listening to Ancestors and Algorithms. If today's episode resonated with you, please leave a review wherever you listen to podcasts and share this episode with a fellow genealogist who is facing this kind of research. That is the best way to help this community grow. Now if you want to go even deeper, the companion guide for this episode includes 12 advanced prompts specifically designed for African American genealogy research, including a multi-step workflow for navigating the eighteen seventy wall using all four tools we covered today. There's also a GPS research checklist tailored to this specific research scenario and a guide to the Freedmen's Bureau record types and what each one contains. It is available to Patreon members at ancestorsai.com. The free episode you heard today stands completely on its own, but the companion guide is where the serious research toolkit lives. For everything you need, every episode, the companion guide library, and our private community, head to ancestorsandai.com. It's all right there waiting for you. 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.