Ep. 38: The Research Map - How AI Finds the Records You're Missing

Most genealogists search 4-6 databases and miss 70% of the records that exist for their ancestor. In this episode, we fix that.

If you have ever searched Ancestry, FamilySearch, and a couple of other databases and still hit a wall, this episode is for you. The problem is not that the records do not exist. The problem is that you do not know what records exist or where to find them before you start searching. That is the gap this episode closes.

In Episode 38, Brian introduces the Research Map: a structured, AI-powered framework you build before you search a single database. Using Claude, ChatGPT, and NotebookLM, you will learn how to map every record type, every repository, and every access pathway relevant to your specific ancestor, including the ones that never appear on Ancestry or FamilySearch.

In this episode, you will learn:

• Why most genealogists are only searching 20-30% of the records that actually exist for their ancestor, including the three mental habits that keep them there
• The Research Matrix Prompt: how to use Claude to generate a comprehensive, prioritized list of every record type relevant to your ancestor's time, place, and background. Results are organized by repository, digitization status, and research priority
• The "What Am I Missing" Prompt: how to use ChatGPT to surface specialty repositories, local archives, ethnic community records, and record types that no major platform indexes. Includes how to flag which suggestions need verification before you act on them
• How to use NotebookLM to cross-reference and synthesize your research map into a source-backed, conflict-resolved action plan
• Why a 1880 Special Schedule of Agriculture for Iowa was sitting on FamilySearch the entire time, searchable for free, and why it never showed up in a standard census search
• How Iowa state censuses going back to the 1840s represent years of uncollected evidence that most researchers have completely skipped
• The difference between a failed search and a genuine absence of records, and why confusing the two stops research in its tracks
• How GPS Element 1 (Reasonably Exhaustive Research) defines the standard professional genealogists use, and how AI helps you meet it

The AI Tools Featured:

• Claude (claude.ai): Research Matrix Prompt, record type mapping, repository identification
• ChatGPT (chatgpt.com): Specialty repository brainstorming, ethnic and local archive surfacing
• NotebookLM (notebooklm.google.com): Source-grounded synthesis and cross-referencing

All workflows use free tiers. No paid subscription required to follow along.

The Genealogical Proof Standard Connection:</b> This episode is Part 3 of the GPS Mini-Series within Ancestors and Algorithms. Episode 38 focuses on GPS Element 1: Reasonably Exhaustive Research, which is the standard that says you must search every source that could reasonably be expected to hold information about your ancestor before drawing a conclusion. AI does not replace that standard. It helps you finally know what that standard requires.

The Teaching Scenario:</b> This episode uses a composite ancestor named Silas Renner, a post-Civil War German-American farmer in Buchanan County, Iowa, with a documented two-year gap in his record between 1865 and 1867. The research map built in this episode surfaces record types most researchers have never searched, and which found three records that had been sitting in free databases the entire time.

For Australian and UK Genealogists:</b> The Research Map framework applies directly to your research. For Australian researchers, the same framework surfaces resources including Trove (trove.nla.gov.au), the National Archives of Australia (naa.gov.au), Public Record Office Victoria, and State Records NSW, covering record categories that sit outside the major platform indexes just as they do for American research. For UK researchers, the framework applies equally to county record offices, the British Newspaper Archive, ScotlandsPeople (scotlandspeople.gov.uk), and specialist collections at The National Archives (nationalarchives.gov.uk). The method is identical. Different archives.

Resources Mentioned:

• FamilySearch Iowa Non-Population Census Schedules (familysearch.org)
• FamilySearch Iowa Grand Army of the Republic Membership Records (familysearch.org)
• Chronicling America (Iowa newspapers) at loc.gov/chroniclingamerica
• HathiTrust Digital Library (county histories) at hathitrust.org
• State Historical Society of Iowa (iowaculture.gov)
• GPS Mini-Series: Episode 30 (Overview), Episode 35 (Element 4), Episode 38 (Element 1), Episode 42 (Element 5, coming)

Companion Guide: The Companion Guide for this episode includes 12 advanced prompts built on the Research Map framework, including specialized versions for African American pre-1870 research, international non-English archives, and DNA-integrated research planning. Available for Patreon members at ancestorsandai.com.

Join the Community: "Ancestors and Algorithms: AI for Genealogy" is our private Facebook group for genealogists learning to use AI tools in their research.

For everything, including every episode, the community, Companion Guides, and The Research Lab: visit 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 0:11

There's a moment in genealogy research where everything changes. Not the big breakthrough moment, not the missing relative found. I'm talking about the moment when you realize you haven't actually been searching. You've been skimming the surface of a very deep ocean. For me, that moment came on a Tuesday afternoon, staring at a record type I had never heard of in fifteen plus years of research. A special schedule. Tucked inside an 1880 census collection for an Iowa farming county, documenting every acre, every bushel of wheat, every head of cattle on a piece of land owned by a man I'd been researching for months. I had searched the census repeatedly. I thought I knew that census cold. What I didn't know was that it had a companion document sitting quietly in a completely separate database, invisible to every search I'd ever run. That afternoon, my understanding of what it means to be thorough changed permanently. Today, I'm going to show you how AI helped me build something I should have built years earlier, a true research map. Not a list of where I've looked, a map of everywhere I should be looking. Let's dive in. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host Brian, and today we're deep in what I've been calling the GPS mini series. If you've been with us for episodes 30 and 35, you know what this is. We're taking the five elements of the genealogical proof standard and showing you exactly how AI tools help you work through each one. Episode 30 was the 30,000 foot review. Episode 35 went deep on element four resolving conflicting evidence. Today we're going back to the beginning. Element one. Reasonably exhaustive research. What does that phrase actually mean in practice? Because I can tell you, most genealogists, myself included for a long time, think we're already doing it. And we're not. Today we fix that. Three AI tools, one powerful framework, and a method you can start using on your most stalled ancestor before the day is out. Let's get started. Let me give you a number. Ready? Most genealogists actively search between four and six databases, and when you add up all the records in those four to six databases, you're likely covering somewhere between twenty and thirty percent of the records that could exist for your ancestor. Not because the other 70% doesn't exist, because you don't know how to look for them. That's not a thorough search. That's a good start. Now I'm not saying that to make you feel bad. I'm saying it because it happened to me. For years my research routine looked something like this ancestry, family search, find my past, maybe a Google search if I was feeling adventurous. I thought I was being comprehensive. I found a lot of records. I felt like I was doing thorough work. What I didn't find was the record I was missing. Not because it didn't exist, because I didn't know how to look for it. The genealogical proof standard, the framework that professional genealogists use to evaluate the quality of their conclusions, calls this out directly. Element one isn't search the obvious places. It says reasonably exhaustive research. And the emphasis is on exhaustive. You're expected to search every source that could reasonably be expected to hold information about this person in this time, in this place. That's a high bar. And most of us don't even know what the bar looks like. Let me introduce you to the composite teaching scenario I'm using today. His name is Silas Renner. Silas was born around 1843 in Perry County, Ohio. He served in the Union Army during the Civil War, somewhere in an Ohio regiment, and by the summer of eighteen sixty seven, he had purchased land in Buchanan County, Iowa, about forty miles north of Cedar Rapids. He was a farmer, he married, he had children, he died in nineteen oh one. And there's a gap. Between eighteen sixty five when he was mustered out of the Army and July eighteen sixty seven, when his name first appears in an Iowa land deed, Silas Renner's record goes quiet. He's not in an Ohio state census for eighteen sixty five because Ohio didn't have a state census. He's not in a federal census because the next one isn't until eighteen seventy. The Iowa state census for eighteen sixty seven might show him if he arrived before the enumerator came through, but it's been underused as a research tool because most researchers don't even know Iowa ran state censuses in those years. Two years, twenty six months, unaccounted for. Here's where the research breakdown happens, and I want you to think about your own research as I describe this. Three traps keep most genealogists operating inside a small search radius. Trap one is the assumption that indexed equals complete. If a record isn't showing up on Ancestry or Family Search's main search interface, many of us assume it either doesn't exist or isn't findable. That's not accurate. Thousands of records are unindexed, microfilmed but not digitized, or sitting in a county courthouse that hasn't partnered with any of the major platforms. Trap two is the assumption that digital equals everything. If you can't find it online, the record doesn't exist. But think about what that means. We are in the middle of the greatest digitization effort in the human history, and we are still nowhere near done. Billions of records exist only on paper, in person, at a courthouse, at a state historical society, at a church archive. Trap three is the assumption that the big platforms have it all. Ancestry, family search, and find my past are extraordinary, but they're not omniscent. There are county historical societies, state archives, newspaper morgues, fraternal organization records, agricultural schedules, church auxiliary records, and local business directories that are not on any major platform and never will be. When you add up these three traps, you get what I call the illusion of thoroughness. You feel like you've searched. You have searched, but you've searched the most visible fraction of what exists. So what do you do about it? You build a research map before you start searching. Not a list of where you've looked, a map of everywhere you should be looking. And that's exactly what I'm going to show you how to do today. That's where AI becomes your research assistant, not your researcher. Our golden rule applies here as much as anywhere. AI can't visit the courthouse for you. It can't read the handwritten deed that was never digitized. But it can do something you and I genuinely struggle to do on our own. It can push past our mental habits and show us territory we didn't know existed. Three tools today Claude first to build a research map, ChatGPT second to push the edges of that map, NoBook LM third to organize everything into a plan we can actually execute. Let's get to work. Before you search for a single record, you need to know everything you should be searching for. This sounds obvious, but almost nobody does it systematically. Most of us go straight to the database, type the name, see what comes back, and search until we run out of ideas. What I want you to do instead is treat this like a briefing before an expedition. Before you enter the field, you need a complete picture of the terrain, every possible record type, every repository that holds material from this time and place, every source that could, in any plausible scenario, document this person's life. Claude is exceptional at building this map. Not because it knows what's currently in the Buchanan County Courthouse, it doesn't, but because it has deeply internalized the structure of American genealogical record keeping across different eras and regions, and it can lay out that structure in a way that would take you days to build from scratch. Here's exactly what I typed. This is the research matrix prompt, and I want you to copy this for your own work. Prompt number one Quote, you are an experienced professional genealogist specializing in 19th century American research. I'm working on a composite teaching scenario for educational purposes, a farmer named Silas Renner, born approximately 1843 in Perry County, Ohio. He served in the Union Army during the Civil War and Ohio Regiment. After the war, he purchased land in Buchanan County, Iowa in July 1867. He farmed in that county until his death in 1901. He was of German American heritage. I have already searched the 1870 and 1880 Federal Census, his death certificate, and his Buchanan County land deed from 1867. I located a Civil War Service Record Index card confirming his regiment but have not yet obtained the full compiled military service record. Please create a comprehensive research matrix with the following columns record type, approximate date range, primary repository, whether it is typically digitized or microfilmed, whether it requires an in-person visit or research request, and your priority rating, high, medium, or low based on likelihood of new information. Please include record types that most genealogists overlook when researching a post-Civil War Midwestern farmer. Be specific about repositories and note any caveats about access or survivability end quote. Now, before I tell you what Claude returned, let me break down why this prompt works. First, I established the expertise level at the top. Experienced professional genealogist specializing in 19th century American research. That framing matters. It primes Claude to answer at a professional level rather than giving you a beginner's overview of census records and obituaries. Second, I gave Silas's complete known biography. The location, the occupation, the heritage, the time period, the more specific context you provide, the more targeted the matrix will be. Iowa Farmer of German Heritage in 1870 generates different repository suggestions than East Coast Merchant in 1820. Third, and this is critical, I explicitly listed what I had already searched. Without this, Claude will spend half its responses telling me to check the censuses and the service record. By flagging what I have, I'm asking it to think beyond my existing search radius. Fourth, I asked for a structured matrix with specific columns. Not a paragraph of suggestions, a table I can actually work with and check off as I go. What did Claude return? An extensive research matrix. Some of what included I expected. Some of it stopped me cold. The expected discoveries, county deed records, which I already had, county probate and estate records, which I hadn't searched, and Civil War pension files, separate from the service record and often far more detailed. The stops you cold discovery, the eighteen eighty Special Schedule of Agriculture. Here's what that is, because I want every listener to hear this clearly. The eighteen eighty Federal Census included supplemental schedules beyond the standard population schedule. One of those schedules documented agricultural production. For every farm of at least three acres with five hundred dollars or more in annual produce value, a census enumerator recorded the acreage, the crops harvested, the livestock count, the value of farm implements. This is not the census you've been searching on Ancestry or Family Search's main interface. It's a separate collection, sometimes separately cataloged, and entirely invisible to a standard census name search on most platforms. The 1880 non-population census schedules for Iowa are available on FamilySearch. If Silas Renner had a working farm in 1880 and his census entry strongly suggests he did, he should appear in that agriculture schedule. I had never looked there. Not because I was careless, but because I didn't know that schedule existed as a distinct resource. Collot's Matrix also flagged something else I hadn't thought of Iowa State Census Records. Iowa ran statewide state censuses at regular intervals, including the 1865, 67, 69, 73, 75, 85, and 1895, among others going back to 1847. The key caveat is that not every county participated in every enumeration year. So the first research task is confirming which years Buchanan County actually appears in before assuming an 1865 or 1867 record exists. The 1867 Iowa State Census, the same year as Silas's land deed, could potentially show him in Buchanan County if he arrived early enough in the year and if Buchanan County participated in that enumeration. I had been treating 1885 and 1895 as the main Iowa State census years because those are the most prominently indexed on FamilySearch's main interface. I had completely missed that Iowa had been running state censuses since the eighteen forties. That matrix also pointed me towards something I find genuinely important for the specific gap, Grand Army of the Republic Post Records. The GAR was the veterans organization for Union soldiers after the Civil War. Iowa had an active department with local posts in nearly every county. Members paid dues, elected officers, recorded deaths and funerals of fellow veterans, and kept meeting notes. For a Civil War veteran who stayed in Buchanan County for thirty plus years, membership in the local GAR posts would be almost expected. Iowa GAR membership records are actually searchable on Family Search in the collection called Iowa Grand Army of the Republic membership records. The records are arranged by county and then by post and include each veteran's name, war record, birth and death information, and family data. The State Historical Society of Iowa and Des Moines holds additional post minutes and materials beyond what's indexed online, but the membership records themselves are accessible without a research request or an in-person visit. One more thing on the Matrix County newspapers. The Buchanan County Bulletin published in Independence, Iowa from eighteen sixty nine to eighteen ninety one, and its predecessor papers covered the county going back to the eighteen fifties. Local papers from that era regularly published farm sale notices, new arrivals column noting who had recently moved into the area, church news and legal notices. Silas could appear in the Buchanan County Bulletin without that appearance being indexed anywhere, and portions of this paper are available through the Chronicling America Database free to anyone with an internet connection at loc.gov. The matrix also flagged township level tax assessment records. When Iowa farmers settled on land, they registered with their township assessor for annual property tax purposes. These tax rolls kept county by county could potentially document Silas' presence in Buchanan County in eighteen sixty seven, eighteen sixty eight, and eighteen sixty nine, the very years between his land need in the eighteen seventy Federal Census where he's invisible. Tax roll for this era are not typically digitized. Many are held at the county level and some may be at the state archives. Their survival is inconsistent, which is why Claude rated them as medium priority rather than high, but they exist. Without the matrix, I wouldn't have thought to ask. One more item from the Matrix that I want to highlight because it catches almost everyone off guard county biographies and local histories. Iowa counties frequently commissioned subscription histories in the eighteen eighties and eighteen nineties, where prominent farmers paid to have short biographical sketches included. If Silas Renner was of sufficient standing in his community after twenty plus years of farming in Buchanan County, he might appear in a county history published around the time of his death. Some of these histories are available through HathiTrust at HATHITRUST.org. Others are on the Internet Archive, and others have been transcribed by genealogy volunteers and posted on county genealogy websites that don't show up in any standard database search. Here's the honest assessment of what Claude gave me. It handed me a research map with six to eight times more territory than the four databases I've been working with. And it did it in about forty seconds. That map is not a guarantee of results. Claude doesn't know whether the Buchanan County 1870s arid tax rolls survived a courthouse flood or fire. It doesn't know whether Silas's German Lutheran congregation kept records in the Signode archive or whether those records are long gone. It gives you the framework you have to verify the details. And I want to address one more thing before we move on, because I've gotten this question from listeners enough times that it's worth naming directly. You might be thinking, why not just use perplexity for this? Perplexity is excellent at current web research with citations. And yes, I use it regularly. But for building a research matrix grounded in historical record types across specific time periods and geographies, Cloud structured reasoning and professional genealogy knowledge is the right tool. Perplexity strength is finding what's out there today. Cloud strength is knowing what should have been created 150 years ago. Different tools for different tasks. That's why the tool to function lock exists. That verification of what's out there today is actually the strength we're about to use in ChatGPT, where the approach is different again. That verification is where tool two comes in. ChatGPT excels at something I'd describe as associative brainstorming. Where Claude gives you a structured, professional, grounded framework, ChatGPT will think sideways in ways that are often exactly what you need. It makes lateral connections between categories. It suggests repositories and record types that a standard genealogical checklist would never include. For Silas Renner's research, I already had a solid matrix from Claude. What I needed now was the second layer, the records that exist outside the standard genealogical playbook, the fraternal organizations, the ethnic community archives, the specialty collections that sit in unexpected places. Here is prompt two. Quote I'm researching a teaching scenario, a farmer named Silas Renner who lived in Buchanan County, Iowa from eighteen sixty seven to his death in nineteen oh one. He was a Civil War veteran of German American heritage who farmed approximately 120 acres. I already have a comprehensive research matrix covering standard genealogical record types. What I need from you is different. I want you to think about local, specialty, and non standard repositories that might hold records about a farmer from this era and region that would not typically be indexed on Ancestry, Family Search, or Find My Pass. Specifically, I'm interested in agriculture and farming. Organization records beyond the federal agricultural schedules, fraternal and veteran organization records beyond standard pension files, ethnic community records given his German American heritage, and any Iowa specific repositories or collections I might not find through national databases. For each category, tell me what the record type is, what information it typically contains, and where such records are most likely held today. And please flag any suggestions you're uncertain about so I know where to verify before I invest research time in that direction. That last instruction, asking ChatGPT to flag uncertainty, is essential. AI tools, including ChatGPT, will occasionally describe specific repositories with great confidence that turn out to be partially inaccurate, combined from multiple real things or in some cases not verifiable at all. Asking the tool to self-flag doesn't guarantee accuracy, but it tells you where to focus your verification effort first. What did ChatGPT return? Some of it was excellent. It suggested grange records. The patrons of Husbandry, the organization behind the Grange Movement, was founded in 1867 and had extremely active Iowa chapters throughout the 1870s and 1880s. If Silas joined his local Grange chapter, which many Midwestern farmers did during that era when agricultural cooperatives were political force, there could be membership roles, meeting minutes, and possibly a death notice. ChatGPT correctly noted that surviving grange records are held inconsistently, sometimes at County Historical Societies, sometimes at the State Historical Society of Iowa, sometimes still with active local Grange chapters. It also flagged German American church records. Specifically, it noted that German settlers in Northeast Iowa frequently organized Lutheran or German Reformed organizations, kept records in German through the eighteen eighties, and that Synod Level Archive sometimes held congregation records separately from the local church itself. If Silas and his family attended a German Lutheran church in Buchanan County, the congregation likely kept a Kirchenbook, a German register with baptism, marriage, and burial entries, possibly in German, and possibly never submitted to any national database. Good lateral thinking. Exactly the kind of record types that fall outside the standard checklist. Then came the dead end. ChatGPT included in its response a description of what it called a Buchanan County Farmers Cooperative Registry, which it described as a county-level collection of farm registration records from the 1870s to the 1890s held at the Buchanan County Historical Society. It explained what information these records contained with complete confidence. I searched for this collection. I checked the Buchanan County Historical Society. I looked through the State Historical Society of Iowa's finding aids. I couldn't confirm this collection exists. This is not a reason to stop using ChatGPT. It's a reason to do exactly what the prompt asked for. Verify before you invest. I've come to treat ChatGPT's specific repository descriptions as hypotheses rather than confirmed facts. Ask yourself, does this sound plausible? Can I find a second source? If yes, add it to your research plan. If no, flag it and move on. ChatGPT's unconfirmed cooperative registry claim went into a verify before pursuing column. The Grange Records and the German Church Records went into the confirmed for investigation column. And that's an appropriate outcome from a brainstorming tool. Now, here's where this gets interesting. I had a Claude Matrix and a ChatGPT brainstorming list, and I needed a way to bring them together and separate what was genuinely from what was still speculation. The tool I reached for was Notebook LM. I'm going to describe what I did here at a high level rather than walk you through the full technique, and I want to be transparent about why. The detailed how-to on using Notebook LM as a source-grounded synthesis tool, specifically the step-by-step workflow for cross-referencing outputs from multiple AI tools and getting citation-backed results, is content I cover in depth inside the research lab episodes one and three of that series, go through the exact prompting approach and the architecture behind why it works the way it does. If you've been curious about what the research lab covers, this is a good example of the kind of depth you'll find there. What I'll tell you is the outcome. I loaded the Claude Matrix, the ChatGPT brainstorming list, a chapter from an 1880s Buchanan County history available through Haythytrust at hatytrust.org, and a State Historical Society of Iowa finding aid into a single notebook LM notebook. I asked it to cross-reference everything, rank what was confirmed by multiple sources, and flag what appeared in only one source and needed verification before I invested research time. The results separated clearly. GAR POST Records, Iowa State Census Records, and County Probate records all appeared in multiple sources and moved to the confirmed list. The Grange records and German Church Styanide records appeared only in the ChatGPT brainstorm and moved to the verify first column. The Buchanan County Farmers Cooperative Registry appeared only in ChatGPT and no other source corroborated it, so it was flagged and set aside. Notebook LM also surfaced two research gaps that none of my assembled sources had adequately addressed. The early Iowa state censuses for 1865 and 1867 without enough repository specific detail, and Silas's specific union regimen which affects which pension records and unit histories are relevant. That cross-referencing is GPS elements three and four working in real time. Analysis and correlation of evidence, resolution of conflicting evidence. I didn't lecture about either one. They appeared as the natural result of doing the research carefully with the right tools. By the end of this pass, I had a research map that covered territory I'd never explored. Now I needed to actually go there. Armed with more than a dozen record categories across eight repositories, I worked through the research matrix systematically over several sessions. What I want to show you is the pattern, because the pattern is the lesson. Every record type that generated new information came from a category I'd never searched before I built the map. Not one. The 1880 Special Schedule of Agriculture for Buchanan County, Iowa. There it was. Silas Renner, Washington Township. The enumerator had recorded 120 acres of improved farmland, 35 acres of woodland, 22 bushels of wheat, 60 bushels of corn, four milk cows, two horses, farm value approximately$900. I had not found this record in 15 plus years of working with the eighteen eighty census. It was there the entire time. The enumerator wrote Silas's name in the same slanted hand that filled every agricultural schedule for that county. It was simply in a different part of the collection, accessible through a different search path on Family Search and invisible to the name search I'd been running. The eighteen eighty five Iowa State Census. I'd been treating the Iowa State Census as something that started in 1885 because those are the years most prominently indexed on major platforms. What the Claude Matric taught me was that Iowa had state censuses going back to the 1840s, with years including 1865, 67, 69, 73, and 1875 before the better known 1885 and 1895 records. That's a meaningful correction to how I'd been thinking about the state's record landscape for years. The 1885 census captured Silas's household in Washington Township with his wife, three older children, and a detail that had never appeared in any federal censuses I'd examined. A fourteen year old nephew listed as a farm laborer. That nephew opened a new line of inquiry into Silas's extended family network and possibly into who might have traveled west with him during the gap years. A GAR Post Record. The State Historical Society of Iowa holds records from Iowa GAR Post partially digitized. I found a transcription of a membership roster from a Buchanan County GAR Post listing, an S Renner joining in the mid 1870s with dues noted through the late 1880s. The description matched Ohio Origin, Civil War veteran. I can't confirm with certainty that this is my Silas without examining the original record, but it's a strong lead that gives me a specific archive to contact for further documentation. Three discoveries, three record types I'd never searched, every one of them appeared in the research matrix, and none of them would have surfaced through the four databases I'd been working with before. I want to stay on the agriculture schedule for a moment because it illustrates something important about why the research map matters so much. The eighteen eighty Special Schedule of Agriculture is not obscure in the sense of being rare or difficult to access. It's available on family search. It's been there for years. But it lives in a separate collection from the main census index, searchable through a different pathway, and it doesn't show up when you type a name into the standard census search. You have to know it exists. You have to know to look for it by geography and then browse it manually or search it through the non-population schedules collection. That's true of dozens of record types. They're not hidden. They're just not indexed in the places where most of us habitually search. The research matrix makes them visible. But the gap the 26 months between 1865 and 1867. I'm going to be honest with you. I didn't close it. Now here is your homework, and I want you to actually do this one. Pick the ancestor you've been stuck on the longest. The one where you've searched everything you know to search and still have the gap. Open Claude right now or say prompt one and use it this weekend. Run the research matrix prompt for that ancestor. List what you've already searched. Ask for 18 or more record categories organized by repository and priority. Then open ChatGPT and run prompt 2. Ask it what you're missing, what the specialty repositories are, what record types fall outside the standard checklist. Ask it to flag its uncertain suggestions. Sit with what comes back. You'll have a combined list of repository leads that is almost certainly longer than anything you've built on your own. You don't have to pursue every item on that list this week. Start with the two or three that are rated high priority and that you've genuinely never searched. That's the minimum return on an hour of effort, and in my experience, it's rarely the minimum. And for those keeping track of where we are in this GPS miniseries, today we demonstrated GPS element one, reasonably exhaustive research as the primary focus. We also put elements three and four into action through the notebook LM synthesis pass, analyzing and correlating evidence from multiple sources and resolving a conflicting claim from the ChatGPT brainstorm. Two episodes remain in this series. Episode 42 will close out the series with element five, the written proof argument. We're building towards something worth the wait. And for my Australian and UK listeners, this exact three tool framework translates directly to your research just with different repositories in the matrix. For Australian researchers, I'd encourage you to run the research matrix prompt with Australian Context. Include Trove at trove.nla.gov.au in your newspaper research, the National Archives of Australia at naa.gov.au for postfederation records, and the state archives for your relevant state, including Public Record Office Victoria, State Records New South Wales, and Queensland State Archives for the kinds of local and specialty records that ChatGPT helped us surface today. For UK researchers, the same matrix approach works beautifully. The National Archives at National Archives.gov.uk, County Record Offices for Local Material, the British Newspaper Archive for the County Newspaper equivalent of what we described, and Scotland's People at ScotlandSpeople.gov.uk if your research crosses the border. Same framework, different archives, every technique from today applies. Thank you so much for listening to Ancestors and Algorithms. If today's episode helped you see your research differently, please take a moment to leave a review wherever you listen to podcasts. And if you know a fellow genealogist who has hit a wall and thinks they've already looked everywhere, send them this episode. Sometimes all it takes is a better map. For my Patreon members, the companion guide for this episode is now in your library. You'll find 12 advanced prompts built around the research matrix framework, including specialized versions for African American pre-1870 research, international research across non-English language archives, and DNA integrated research planning. There's also a multi-tool workflow that chains all three of today's tools together in a single coordinated research session and a GPS research checklist built specifically around Element One. And if you've been thinking about joining, I host a monthly live Q ⁇ A on YouTube exclusively for members and we'd love to have you in the room. 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.