Ep. 45: The Name on the Draft Card

A single 1917 WWI draft card, plus a naturalization petition filed seven years later, reveal an Italian immigrant's real birth name. This is a real, start-to-finish example of using AI for genealogy research, ChatGPT, Claude, and NotebookLM working together to solve it.

If you've searched "AI genealogy tools," "ChatGPT for family history," or "how to use AI to research my ancestors" and wondered where to start, this episode is a full case study, not just a list of tips.

This episode is for you if:

• Your family tree has a name that stops dead, no ship, no naturalization record, no story past "he arrived already American"
• You've heard about using ChatGPT, Claude, or NotebookLM for genealogy but haven't seen a real research case worked start to finish
• You've searched FamilySearch and come up empty, and want to know what a zero-result search actually means
• You're researching Italian ancestry and want to understand how and why immigrant names changed in America

What you'll learn:

• How to use ChatGPT to generate a ranked list of likely name-Americanization patterns, instead of guessing at a single translation
• Why NotebookLM's source-grounded, citation-backed answers can flag a real gap in your own family's paper trail
• How to pivot a genealogy search from a dead-end name to a birthdate-and-birthplace search when FamilySearch's Full-Text Search returns nothing
• How Claude compares two historical documents filed under two different names and states an honest confidence level, not a guess
• Why a WWI draft registration card is one of the most specific, most overlooked genealogy records for cracking identity mysteries

Australian and UK listeners: this same AI-assisted approach works on Australia's WWI attestation papers through the National Archives of Australia and on naturalization and denization records through The National Archives, both of which ask many of the same identity-cracking questions this draft card did.

The honest outcome: this is a Full Breakthrough. The mystery resolves completely, with both identities confirmed on a single naturalization document, using the Genealogical Proof Standard's own methodology, Exhaustive Research, Analysis and Correlation, and Resolution of Conflicting Evidence, done with AI as the research assistant, never the researcher.

Want to go deeper? The Companion Guide for this episode includes advanced ChatGPT, Claude, NotebookLM, Gemini, and Perplexity prompts for building a full name-Americanization search list, a complete multi-record pivot workflow, and a GPS Research Checklist built specifically for identity and name-change genealogy cases. Find it, along with every episode of Ancestors and Algorithms, our AI genealogy podcast, 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

So picture this. A family owns exactly one photograph of him. He's standing in front of a brick building in a work coat, and on the back in pencil, somebody has written a name Patrick Lombard. That's the whole inheritance. No passenger list under that name. No naturalization file under that name. As far as this family has ever been able to tell, Patrick Lombard turns up in Detroit sometime before the First World War, already speaking English, already punching a clock of the Ford Motor Company, with no country behind him at all. Today I'm going to show you how one government form, filled out on a single afternoon in June of nineteen seventeen, cracked open a mystery this family had quietly given up on solving. Let's dive in. Welcome to Ancestors and Algorithms where family history meets artificial intelligence. I'm your host Brian, and today we're tracing an Italian immigrant through the one stretch of American history where the government asked him more specific, more personal questions than almost anywhere else in his paper trail, and then using AI to figure out what he wasn't telling anyone. Now, if you have a name in your tree that stops cold, no ship, no naturalization, nothing before a certain date, like the person arrived already midlife and already American, stay with me. This episode is built for exactly that wall. So let's get started. Here's everything this family actually had before we started. A death certificate for a Patrick Lombard, born about 1888, died in Detroit. A Ford Motor Company retirement pin. A World War II era ration book with the same name. And that photograph, work coat, brick wall, penciled name on the back. Nothing else. No wedding record that named his parents. No baptismal certificate. No letter home. His great granddaughter grew up being told the family was Italian from somewhere down south and that was the entire sentence. Not a region, not a village, only down south the way you describe a direction instead of a place someone was actually born. Any genealogist who's been at this for a while will recognize the shape of this particular wall immediately. It's different from the brick wall where a record never existed at all. This is the wall where the record trail is actually intact, complete even, but it's answering a question you haven't figured out how to ask yet, because you're searching under the wrong name entirely without knowing it. Here's the thing about a name like Patrick Lombard. It doesn't sound like a clue. It sounds like the answer already. Patrick reads Irish. Lombard reads like half the phone book, and if you're picturing Vince Lombardi right now, you're not far off, since it's the same root surname, worn down a little differently over time. If you didn't know better, you'd never go looking for Italy at all, and that is exactly the trap. A genuinely Italian name would have sent the family looking in the right direction from day one. A blank wall would have at least announced itself as a wall. What this family got instead was a name confident enough to stop the surge before it started. We know what doesn't exist, and that's almost more useful than what does. There's no Ellis Island manifest for a Patrick Lombard matching this birth year. There's no naturalization record under that name in Michigan. And here's a detail we're sitting with for a second, because I hear this exact confusion constantly. For a long time, people believed immigration officials changed names at Ellis Island itself, some overworked clerk mishearing an accent and writing down whatever came out easiest instead. That never happened. Passenger manifests were compiled by the shipping companies in Europe before the boat ever left the dock, and the officials at the port were checking travelers against a list, not inventing new names for them. If a name changed and this family's name absolutely changed, it happened later in America on purpose by somebody's own choice. That distinction matters because it means the man in this photograph made an active decision, and decisions leave better trails than accidents do. So here's the human want underneath all of this, and I think it's worth naming plainly before we touch a single record. A twenty-year-old man got on a boat with an Italian name that had belonged to his father and his father's father, and by the time he was in his late twenties, he was answering to something else entirely, five days a week for a paycheck. That's not nothing. That's a trade. He wanted work. He wanted to not be singled out on a factory floor where dozens of languages were already colliding. He probably wanted, more than anything, to get through the next ten years without anyone deciding he didn't belong here. And his great granddaughter, three generations later, wants the thing he traded away. She wants his actual name back and the village that came with it, the specific patch of ground his family stood on before any of this started. I know some of you have been exactly here. Not with this specific man, but with a specific shape of problem. A wall that isn't really a brick wall. It's a name so ordinary it never occurred to anyone to question it. We've all got at least one ancestor whose name feels a little too smooth, a little too easy, and we never stop to ask why, because a smooth name doesn't announce itself as a problem. It sits there quietly, closing a door nobody notices is shut. Detroit in this era helps explain the why here. The city's Italian population sat around 900 people in 1904 and exploded to something like 42,000 by 1925, and the reason was almost entirely mechanical, cars. Most of Detroit's early Italian arrivals settled on the city's east side along Gratio and Harper Avenues, in a neighborhood the immigration themselves nicknamed Caca Lupo, a half jokingly mangling of car loop, after the streetcar turnaround that anchored the area. This wasn't one region of Italy dominating the whole community either. Sicilians, Genoese, and people from Lombardy all landed in the same few blocks, which is worth knowing because it means a Detroit Italian surname doesn't automatically tell you which part of Italy to search. Most of these families arrived through chain migration, one relative or pesano sending word back that there was work, and then a whole village would empty out toward the same three or four American blocks over the following decade. That pattern turns out to matter later in this story. So here's what I did, and I want you to follow along, because this exact approach will work for your own research. I didn't start by trying to guess his real name. I started by asking AI to tell me, honestly, what the existing documents already proved, as opposed to what the family had always assumed. And that gets us to today's mystery question. The only time I'm going to ask it out loud. What was his name before America renamed him? Remember, and I'll say this once now and come back to it later, AI is your research assistant, not your researcher. It wasn't going to hand me his real name out of nowhere. What it was actually good at was something quieter and honestly more useful. I want to start with Nobook LM, and here's exactly why. Notebook LM doesn't answer from the open internet. It answers only from documents you hand it directly, and every answer comes back with a citation pointing to the exact source and passage it pulled it from. That matters enormously at the start of a case like this one, where the biggest risk isn't a missing fact, it's a false one. Family memory tends to smooth things over. Down south becomes a fact nobody questions. I needed a tool that would refuse to do that and would instead show its work. As of July 2026, Notebook LM's free tier gives you a hundred notebooks, fifty sources in each one, and fifty questions a day running on Google's Gemini 3 model underneath. That's genuinely enough for a case like this. I uploaded everything the family had, a scan of the photograph and its caption, the death certificate, the retirement pins inscription, the ration book, and a typed family group sheet somebody put together back in the 1970s. Five sources, all of it text or scanned images NoBook LM could read directly. Here's exactly what I asked it. Quote I've uploaded everything my family has on a man named Patrick Lombard, born about eighteen eighty eight, died in Detroit. Based only on these sources, what is the earliest document that gives a specific, dated fact about him and what information about his origin, if any, appears anywhere in this set? Please flag anything that looks like it might be a genuine gap in the documentation as opposed to something merely unknown to the family, end quote. Nobook LM's answer was more useful than I expected. It pointed out, correctly, that every single source in the notebook postdated 1917, and that nothing in any of the five documents said anything more specific about his origin than Italy, full stop, no region, no village, no ship. It flagged that gap directly. No source in this set documents his arrival, and no source documents a naturalization proceeding, which for a foreign born man who apparently lived his whole adult life in the same American city is, in Notebook LM's words, an unusual silence rather than an ordinary one. That's the moment the case actually opened up. Not because AI found a hidden fact, but because it drew a hard, honest line around what the family's own paper actually said, and that line had a suspicious shape to it. Everything starts at nineteen seventeen. Nothing exists before it. I clicked the citation on that last claim, the way Nobook LM lets you do with every single answer it gives, and it dropped me straight onto the exact line of the nineteen seventies family group sheet where someone had written arrived no record decades ago, as if they'd already suspected the gap and never chased it down. A quick note here for my listeners overseas, because this kind of everything starts at one specific government form pattern isn't unique to America. If your ancestor served or was old enough to be asked to serve, Australia's AIF estation papers through the National Archives of Australia and the UK's World War I service and metal records through the National Archives do very similar work, forcing a specific date, birthplace, and next of kin onto paper for someone who might otherwise have left almost nothing behind. So here's what actually pulled so many Italian immigrants specifically toward Ford. In January of nineteen fourteen, Ford announced what became known as the five dollar day, more than doubling the standard wage from about two dollars thirty four cents for a nine hour shift to five dollars for eight hours. Ten thousand men reportedly showed up at the Highland Park plant the very next morning. Ford also opened the Ford English School that same year, taught by volunteer foremen and clerks, and by 1916 it had grown from 20 students and one instructor to more than 2,000 students and roughly 150 instructors. A diploma from that school could be used toward the requirements for citizenship. The company's sociological department, meanwhile, actually investigated workers' private lives to decide who qualified for the higher wage, which tells you something about how much pressure these men were under to look, sound, and behave like the America Ford wanted them to represent. Even the classroom pushed toward assimilation on purpose. Students learned American geography right alongside their vocabulary, specifically to help them pass the citizenship exam waiting at the end of the road. I think that's where Patrick Lombard comes from. Not a clerk at a port who couldn't spell his name. A shop floor, a paycheck, and a school built specifically to turn immigrant workers into something that read as American on paper. That context matters for what comes next, because it's exactly the kind of background I'd want any AI tool to have loaded before I ask it to reason about a document from this era. Knowing the machinery behind a name change changes how you read the paperwork that survived it. Here's where the actual record hunting starts, and I want to walk through this slowly because the sequence matters as much as the discovery does. The first genuinely new document I found was his World War I draft registration card, the one from the first registration wave, dated June 5th, 1917, for men between 21 and 31. This card asked for a lot. Full name, exact birthday, birthplace, current occupation, employer's name and location, a description of the nearest relative and their address, and a citizenship question with four checkboxes natural born citizen, naturalized citizen, alien, or have you declared your intention? Patrick Lombard's card listed his employer as the Ford Motor Company, his birthplace as Terassini, Sicily, and it had that fourth box checked. He declared his intention to become a citizen. One more wrinkle worth knowing, because it trips people up constantly. DraftBoard sometimes recorded recent Italian immigrants with their surname written first, following Italian convention, which means a card can end up filed in an index under the wrong half of the name entirely. It's worth checking both directions before you conclude a record doesn't exist at all. That single checkbox is worth pausing on because it's a promise that a second document exists somewhere. If Hid declared his intention, there was a declaration of intention on file, a real piece of paper filed at a real court under whatever name he used that day. So I turned to ChatGPT next, and here's why this tool specifically. I didn't need analysis yet. I needed lateral brainstorming, a wide net of plausible name variants, because I had no idea yet whether Patrick Lombard was close to his birth name or nothing like it at all. ChatGPT strength is exactly this kind of associative thinking, generating options a single track search would never surface. As of July 2026, that means free access to GPT 5.5, OpenAI's current model, though you're working inside a rate limit that resets every few hours on the free tier, which was plenty for what I needed. Here's the prompt. Quote I'm researching an Italian immigrant to Detroit, Michigan, who was using the name Patrick Lombard by 1917 and was likely born in Sicily around 1888. Generate a list of Italian given names and surnames that Italian immigrants in this era commonly Americanized into something close to Patrick and Lombard and briefly explain the reasoning behind each pairing. I want a search list to work through, not a single guess, end quote. ChatGPT came back with a genuinely useful list. Pasquale was at the top of it, explained as a very well documented Americanization to Patrick or the nickname Patsy, based purely on sound, nothing to do with Ireland at all. It also suggested Patrizio as a more direct match and flagged Lombardi and Lombardo as the obvious Italian surnames behind an Americanized Lombard, since Italian surnames ending in a vowel were routinely trimmed down when clerks, foremen, or the immigrants themselves simplified them for American paperwork. Now here's the dead end, and I want to be honest about it, because this is exactly the kind of moment where AI can quietly mislead you if you don't verify. I took that list straight to Family Search's full tech search tool. It's entirely free, no subscription anywhere in this workflow. And as of July 2026, it has grown to cover more than 6,600 searchable record collections with well over a billion transcribed results, expanding into naturalization records collection by collection rather than all at once. I searched Michigan's Eastern District Naturalization Records directly for Patrick Lombard. Nothing. Not a false match, not a near miss. Nothing at all. There are two honest explanations for a zero result search like that, and it's worth knowing both before you panic. One is that the collection hasn't been transcribed for full text search yet, since Family Search is rolling this out state by state and record type by record type. The other is that that name I searched genuinely never touched that collection at all. I couldn't tell which one I was looking at yet, and that uncertainty is exactly why the next step matters. That's the moment I had to stop and think instead of searching harder. And this is exactly a verification moment worth naming directly. AI is your research assistant, not your researcher. ChatGPT gave me a smart list of possibilities, but a smart list isn't proof, and searching a good guess is still, at the end of the day, a guess. I couldn't find Patrick Lombard because Patrick Lombard never walked into a courthouse and declared his intention. Pasquale Lombardi did. So I pivoted, and here's the pivot exactly. Instead of searching by name at all, I went back to the draft card and pulled the two pieces of information a namechange can't touch, his birthdate and his birthplace, Terracini, Sicily. This is the same instinct behind the vowel dropping pattern ChatGPT flagged earlier. A recognizable historical pattern is how these names actually moved from one language to another, and it's what told me to stop trusting the name on the record and start trusting the biography underneath it instead. I searched Michigan's Declaration of Intention collection for the Eastern District, which happens to run from 1911 to 1930, for an Italian-born filer with the specific birth date and the specific village. One result came back Pasquale Lombardi Declaration filed in 1913, birthplace Terracini, birth date matching the draft card exactly. Before I got any further, I pulled the actual image of that declaration, not only the index entry, because an index is somebody's transcription and a transcription can be wrong. The document itself confirmed everything. Pasquale Lombardi, occupation listed as laborer, an arrival year matching what little the family had guessed, and a signature that, if I'm honest, I can't prove is the same hand as the one on the 1917 draft card. Handwriting comparison isn't something I trust myself to do reliably, and it's not something I'd hand to AI to certify either. Not for something this consequential. What I could trust were the facts written on the page, and those lined up exactly. I want to slow down here because this is the part where I fed both documents into Claude, and this is exactly why Claude is the right tool for this specific job. Claude's strength is holding multiple documents at once and reasoning across them for consistency, which is precisely what proving these two paper trails are the same human being requires. I gave it the context first. Here's how naturalization law worked in this era. Here's what a declaration of intention versus a petition versus a draft card actually record. Then I asked it to do the comparison. Quote I have two documents about the same research question. Document one is a nineteen seventeen World War I draft registration card for Patrick Lombard, born about March 1888, birthplace Terracini, Sicily, employer Ford Motor Company, nearest relative listed as his father residing at Terracini, and a checked box confirming he had declared his intention to naturalize. Document two is a nineteen thirteen declaration of intention for Pasquale Lombardi, same birth month and year, same birthplace, Terracini, Sicily. Walk through whether these two documents plausibly describe the same person, state your confidence level, and identify which specific data points are doing the most work in that conclusion, end quote. Claude's answer laid out its reasoning step by step. Same birth month and year is suggestive but not proof on its own, since two men from a small immigrant sending region could share a birth year by coincidence. What tipped it toward high confidence was the shared specific village, Terracini, a town small enough that two unrelated Italian immigrants landing in the same American city with the same birth month from the same small Sicilian town was statistically unlikely to be a coincidence. It also noted correctly that this was still correlation, not certified proof, and that a document naming both identities directly would settle it completely. That last line was basically a dare. So I went looking for exactly that, a document naming both identities directly. Here's what a petition for naturalization actually is, because it matters enormously for what's coming. After filing a declaration of intention, an immigrant generally had to wait at least two years and had to have lived in the United States for a minimum of five years total before filing the petition, the second and final step towards citizenship. Petitions and declarations didn't even have to be filed at the same courthouse, which is worth knowing if you ever hit a similar wall in your own research. And here's the detail that cracks this whole case wide open. Starting in 1906, a petitioner could formally request, right there in the petition itself, that the court legally change their name as part of the naturalization decree. Most people made that request because by the time they filed their petition, they'd usually been living under an American name for years already, on paychecks, on leases at church. The petition let them make it official in front of a judge with two witnesses swearing to who he'd become. I searched the Eastern District of Michigan's naturalization petitions using the same specific details, the 1913 declaration date and Terracini as birthplace, and I found it. A petition for naturalization filed in 1920, seven years after the declaration, right at the edge of the legal window before a declaration expired. The petitioner's name is printed as Pasquale Lombardi, and directly beneath it in the same clerk's hand, in the same ink, is a line most petitions from this era carried almost as a boilerplate, except this time it wasn't boilerplate at all, also known as Patrick Lombard, with the court's order right beneath it, granting the change and recognizing the new name going forward. Two names, one man, one document, one moment permanently on paper. I want to sit with that for a second before I explain anything else about it. For years this man lived two lives on two different sets of paper, Pasquale Lombardi at the courthouse where his original declaration lived, Patrick Lombard everywhere else, the time card, the draft board, the neighbors. And then, on one afternoon in 1920, a clerk wrote both of those names onto the same line and closed the gap between them for good. He wasn't hiding. He was becoming, on purpose in installments, and this is the receipt. So here's what the evidence actually tells us plainly. Patrick Lombard was born Pasquale Lombardi around March of eighteen eighty eight in Terracini, a coastal town in the province of Palermo, Sicily. He arrived in the United States sometime before 1910, settled in Detroit's East Side Italian community, filed his declaration of intention under his birth name in 1913, went to work for Ford Motor Company sometime around the $5 day hiring surge of 1914, registered for the draft under his Americanized name in 1917, and formally, legally became Patrick Lombard in 1920, the same year he became a citizen. His great granddaughter now has both things she wanted. She has the name her family actually carried out of Sicily and she has the village, which means church records, civil registration, and possibly living relatives in Terracini are all now genuinely searchable in a way they weren't twenty four hours ago. Sicilian parish baptismal registers routinely run back well into the seventeen hundreds, so this isn't the end of the search, it's the actual beginning of it. The part the family had assumed was permanently lost. That's not a small thing. That's a door that's been shut for four generations and it's standing open now. So quick recap. A name that sounded too American to question turned out to be the entire mystery. A draft card gave us a birthplace and a checkbox that promised a second document existed. AI didn't hand us the answer at any point, but it organized the gaps honestly, widened our search net, and helped us hold two documents up against each other until the pattern became undeniable. And a naturalization petition, of all things, turned out to be where a man's whole story finally got written down in one place. Here's your homework this week. Go find the earliest solidly dated document for every surname line in your tree and ask yourself honestly, is there a hard floor there? A point which absolutely nothing exists? If there is, go looking for a World War I draft card. Even for a man you're sure never served, because registration was mandatory paperwork, not enlistment, and virtually every American man of the right age filled one out whether he saw a battlefield or not. That one form asks more nosy, specific questions than almost anything else on record. Exact birthplace, employer, a physical description, and the name of the one relative he trusted enough to list. Pull the birthdate and the birthplace off of it, set the name itself aside for a moment, and search by everything else instead. That's the move that cracked this case, and there's a good chance it's sitting in your own tree waiting for the same treatment. And for my Australian and UK listeners, the techniques we cover today translate directly to your own research. If you're chasing a name that seems to appeared out of nowhere, Australia's World War I attestation papers held by the National Archives of Australia ask many of the same specific questions that this draft card did, next of kin, birthplace, occupation, and are searchable online today. UK researchers will find similar value in service and metal records held by the National Archives, alongside naturalization and denization records that, much like this Michigan petition, sometimes record a legal name change directly. Same approach, different archives. Thank you so much for listening to Ancestors and Algorithms. If you enjoyed this episode, please leave a review wherever you listen to podcasts. And if you know a fellow genealogist who could benefit from what we cover today, share this episode with them. That is the best way to help our community grow. If today resonated, the companion guide takes it a step further. I include the exact multi-record search strategy I use to pivot from a dead-end name search to a birthdate and birthplace search, plus advanced prompts for building out a full name Americanization search list with AI and a GPS research checklist built specifically around identity and name change cases like this one. The free episode stands on its own, but the companion guide is where the full toolkit lives. For everything you need, including every episode, our private Facebook community, companion guides, and the research lab, head over to ancestorsnai.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!