Ep. 46: The Paper Son - Chinese Roots and AI | Cracking a Chinese Exclusion Act Case

During the Chinese Exclusion Act, one immigrant's 200-page federal case file was built to hide his real name, and nearly succeeded.

For sixty-one years, the Chinese Exclusion Act barred an entire ethnicity from immigrating to the United States, with one narrow exception: the children of Chinese American citizens. That loophole created the "paper son" system, families with no blood relation to a citizen buying a purchased identity, memorizing a stranger's village and household, and surviving a formal cross-examination at Angel Island to get through the door. In this episode, we follow one of those case files: two hundred pages of coached interrogation testimony, hand-drawn maps, and a personal letter in Chinese that the government never meant to be evidence of anything except fraud, and that turned out to hold the truth instead.

This episode is for you if you're researching Chinese American genealogy, tracing an ancestor from the Chinese Exclusion Act era (1882 to 1943), working with Angel Island immigration case files, searching for a "paper son" or "paper daughter" identity in your family tree, dealing with Cantonese or Toishanese surname romanization, or trying to translate a handwritten document in a language you don't read yourself. We show our work with three AI tools, each doing a specific job. Claude builds a side-by-side comparison table across two separately-taken interrogation testimonies and flags exactly where a coached story starts to crack under pressure. Gemini, through the free AI Studio, transcribes and translates a handwritten Chinese-language letter, the piece of the file that finally answers who this man really was. Perplexity, through the free Comet browser, tracks down the historical context and points us to a real, free resource for identifying an ancestral Chinese village: the Friends of Roots Village Database.

For my Australian and UK listeners, this record type has real cousins in your own archives. Australia ran its own version of exclusion through the Immigration Restriction Act of 1901, and Chinese residents needed a Certificate Exempting from Dictation Test just to leave the country and come home again, records held today by the National Archives of Australia. In the UK, the parallel runs through the Aliens Act of 1905 and the Home Office alien registration files held at The National Archives in Kew. Same instinct, different door, same kind of file left behind.

Honest outcome: this is an Unexpected Discovery. We find the real answer, but it is not the answer four generations of one family believed about themselves. The Chinese Exclusion Act was state racism enacted as law, and this episode treats that history, and the resourcefulness it took to survive it, with the weight it deserves, not as a plot twist.

Want to go deeper than the free episode? The Companion Guide includes twelve advanced AI prompts and full workflows for this exact kind of case: spotting a coached interrogation, verifying a translation you can't read yourself, and building a properly cited research file. Patreon members also get five intermediate prompts built around real, verified record collections for Chinese Exclusion Act research, including where files are actually held, how to search around a broken index, and how to tell the National Archives from USCIS before you send a request to the wrong agency. Find all of it, and every episode, 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 0:11

Here's a fact that stopped me cold when I found it. Somewhere in a federal record center in San Francisco, there's a file on one immigrant that runs close to 200 pages. Sworn testimony, a hand-drawn map of a house on the other side of the Pacific, photographs, even a letter written in Chinese that nobody left in the family could read anymore. Now here's the strange part. His own younger brother came through the same port decades later and left behind a file you could read over breakfast. Two pages, maybe three. Same family, same ocean. So why did the government need two hundred pages to let one man in and almost nothing for the other? Let's dive in. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host Brian, and today we're heading into one of the richest and honestly one of the most painful chapters in American immigration paperwork, the Chinese Exclusion Act, and the families who found a way through it anyway. Now, maybe nobody in your tree ever set foot on Angel Island. Stay with me anyway. The record type at the center of today's story, a government file built entirely out of suspicion, shows up anywhere a country decided one group of people needed extra watching. Learn how to read one of these files, and you'll know how to read all of them. So let's get started. Let me introduce you to a young man the family has always called Ing Yuk Sing. That's a composite name, and he's a composite person. I built him from patterns I found across dozens of real exclusion era case files, and I want to be upfront about that before we go any further. But every single detail I'm about to give you happened to real families over and over in exactly this shape. He was born around 1904, somewhere in the Toishan region of Guangdong province, the part of China that sent more people to Gongsein, Gold Mountain, meaning America, than almost anywhere else. His birth family had a little land and a lot of mouths to feed. He was the oldest son, and when a chance came along to send him to America, his family found the money. Not through a normal application, through something else entirely. Here's the piece of history that makes this story possible. In eighteen eighty two, the United States passed the Chinese Exclusion Act. It's worth naming plainly what the law was. It barred an entire ethnicity of people from immigrating based purely on race, with a narrow list of exceptions for merchants, teachers, and a few other categories. It stayed on the books, in one form or another, for 61 years. Then, in 1906, the San Francisco earthquake and fire tore through the city hall and destroyed the birth record stored inside it. Suddenly, any Chinese resident of San Francisco could claim to have been born there. No paper existed anymore to prove otherwise, one way or the other. Some of those men really were born there. Some weren't. Either way, a birth in San Francisco meant citizenship, and citizenship meant the right to bring over children from China, one of the only doors the exclusion act left open. That door had a price. Families in China who had no blood connection to a citizen at all could pay for a slot as his son. The buyer memorized a coaching book full of village details, family names, and household layout, all of it belonging to someone else's life. Genealogists call these men paper sons. It's a term I want us to sit with for a second because it wasn't a metaphor to the people living it. It was survival dressed up as somebody else's family. That's the slot our young man's family bought. A merchant in Sacramento, a man we'll call Ing Bak Jung, needed a son on paper. The broker who arranged it wrote up a coaching book, pages of facts about the Eng household that the boy had never seen with his own eyes. How many rooms? Which direction the door faced? The names of neighbors he would never meet. He studied that book the way another boy might study for an exam that decided everything. And so at seventeen he became Ing Yu Xing. He memorized a life that wasn't his and boarded a ship. It was bound for a country that had spent almost forty years by that point writing laws to keep people like him out of it. I want to name what he actually wanted underneath all of that. He wasn't chasing an adventure. He was the oldest boy in a family that couldn't feed itself, and he was being sent across an ocean to fix that, carrying a stranger's biography as his only ticket in. That's not a small thing to ask of a seventeen year old. So here's where I think a lot of you are already nodding, even if nobody in your family ever touched Chinese exclusion. Almost everyone doing this kind of research has hit a name that doesn't sit right. A surname that shifts between census years, a birthplace that contradicts itself from one document to the next. Usually we chalk it up to a tired clerk or bad handwriting. Sometimes, though, the contradiction is the whole story. Sometimes the family changed the truth on purpose because the truth wasn't going to get them through the door. That's exactly what happened here. And it's why, generations later, Yuk Sing's own descendants are left holding a name, Eng, that their own family history society has no matching lineage for. They know the name their great grandfather used every day of his American life. They don't know if it was ever really his. If you've ever felt that particular flavor of doubt, the one where a document you've trusted for years suddenly looks like it might be hiding something, you already know the feeling this family lives with. Most of us get one or two documents like that in an entire research career. This family got an entire identity built on one. So here's the question that opens this whole case, and I'm only going to ask it once. Why does this one man have close to 200 pages of federal paperwork and his own brother almost none? To even start answering that, I needed AI. This is where I bring in the golden rule that guides everything on this show. AI is your research assistant, not your researcher. Claude was about to do some heavy lifting on a transcript longer than most short novels. But it couldn't tell me what was true. Only I could do that against the actual documents. Let me show you how that split worked in practice. The case file itself came from the National Archives at San Francisco. The office holds more than 200,000 of these investigation files from the exclusion era, covering people who passed through San Francisco and Honolulu alike. Yuk Sing's file included the standard pieces you'll find across most of them, a certificate of identity, photographs, a hand drawn map of the family's house and village, correspondence, and the interrogation itself. Some files in this collection run a few pages. Others, like this one, run for hundreds, and the length itself turns out to be a clue worth paying attention to before you read a single page of testimony. The interrogation is the heart of the file, and it's brutal in its detail. When Yuk Sing arrived at Angel Island in nineteen twenty one, he didn't face one friendly officer with a checklist. He faced a board of special inquiry, two immigration inspectors, a stenographer and an interpreter, all cross examining him and his alleged father separately on the same set of facts. How many houses stood in the village? Which one was theirs? Who lived to the left? Who lived to the right? How many pigs the neighbors kept? Every answer had to match the other man's answer, given in a different room, with no chance to compare notes first. He was detained for weeks while officials worked through it, held in a barracks building on the island with dozens of other men waiting on the same verdict. Some of those men, in barracks that still stand today, carved poems into the wooden walls while they waited. Close to two hundred of those poems survive right now, preserved as part of the historic site. They weren't writing about paperwork. They were writing about homesickness and fear and how strange it was to be judged worthy or unworthy of a country from inside a locked room on an island in its harbor. Yuk Singh's own weight produced no poem that survived, as far as anyone knows. It produced a transcript instead. And that back and forth, spread across multiple interrogation sessions, is exactly why his file runs to nearly 200 pages. Now I had a 40 page transcript of that interrogation, testimony from both father and son, and I needed to know where their stories actually lined up and where they didn't. Reading 40 pages of question and answer testimony start to finish, holding every detail in your head at once, is close to impossible for a human brain. You get to page thirty and forget exactly what page four said about the well. This is where Claude earns its place in the workflow. I didn't reach for it because it's the newest thing or the most talked about thing. I reached for it because document analysis and pattern recognition across a long, repetitive transcript is exactly the kind of task it handles well. And because I could keep the whole case building inside one ongoing project instead of starting from zero every time I opened a new chat. Here's exactly what I typed. Quote, I'm going to give you the interrogation transcript from a 1921 Chinese Exclusion Act case file. Two people were questioned separately about the same family in the same village in China, the young applicant and the older man claiming to be his father. Please build a simple table. One row per topic, number of siblings, house layout, neighbors, village landmarks. One column for each person's answer. Flag any row where the two answers don't match. Don't summarize or interpret anything. Extract only what each person actually said word for word where you can end quote. What came back was a clean, sortable table, twenty some rows deep, most of them match perfectly. That's not surprising. Coaching worked most of the time, or these files wouldn't exist in the thousands that they do. But three rows didn't match. A number of siblings that shifted by one between the two testimonies, a description of the house's roof that differed in a small but specific way, and a detail about a neighboring family's livestock where the father hedged two, maybe three, where the son had answered flatly two. Small cracks not proof of anything yet, but cracks are where a genealogist starts digging, and I want to be honest that Claude found them in about ten minutes flat. It would have taken me an entire evening to build that same table by hand, cross checking pages against each other, keeping every prior answer straight in my head while I read the next one. That's the whole value of the tool here. It's fast at comparison and it doesn't get tired on page thirty the way I do. It's not the one deciding what those three cracks mean, and it never pretended otherwise when I asked it directly whether the mismatches proved anything. It told me, correctly, that inconsistency alone doesn't prove a false claim. People misremember real details too. That part, the actual judgment call, was still coming, and it took two more tools to get there. Three small inconsistencies in a coached interrogation could mean almost nothing. People misremember details under pressure, even details they rehearsed. So before I went any further, I wanted historical context. Was this pattern of small mismatches typical or was something else going on with this particular file? I opened perplexity through the free comment browser and gave it a research question rather than a request for an answer. Quote I'm researching a Chinese immigrant who arrived in San Francisco in 1921 as an alleged paper son. Search the National Archives finding AIDS and Chinese American Genealogy Society resources for how families connected to Toishan District, Guangdong Province are typically documented, including any family or district association records that might help trace a village of origin. Give me sources I can check myself with links, end quote. Perplexity came back with something I hadn't expected to need yet, an explanation of Chinese district and family associations. These were groups that Chinese immigrants organized by shared home district, and some of them kept their own genealogies, called Jiapu, alongside cemetery and membership records. It also flagged something small but important. In records from this era, you'll sometimes see a name prefixed with a as in a sing. That's not a surname. It's closer to how we'd say he sing, a familiar nickname, and generations of American clerks wrote it down as if it were a legal part of the name. I mentioned that because it's a trap that catches researchers constantly, and it's worth knowing before you ever open one of these files. It also gave me something I didn't know I needed, a reminder that romanization itself is a moving target here. The same Chinese surname can turn into a half a dozen different English spellings depending on which dialect the clerk thought he was hearing, and Toyshan District spoke its own distinct dialect within Catanese. A name spelled one way on a ship manifest and a slightly different way on a case file isn't necessarily two different people or a lie. It might be nothing more than two different clerks on two different days guessing at the same sound in two different ways. One resource perplexity pointed me to is worth naming directly because you can use it yourself this afternoon. It's called the Friends of Roots Village Database. It's a free tool built from decades old immigration registers that lets you search Toishan, Sun Lui, Huangping, and Chungshan counties by surname and pull up the actual villages associated with that name. It won't hand you an answer. It will hand you a short list of real candidates instead of an entire province to search blind. That's the habit good genealogists build over years, casting a wide enough net to know what kind of records could even exist before assuming they don't. None of that told me where his real village was, though. For that, I turned to the one piece of the file I'd been avoiding. It was in Chinese, and I don't read a word of it. A personal letter, three pages, folded into the back of the case file. Officials seized and reviewed correspondence like this routinely, looking for anything that undercut a family story. This particular letter had done exactly that to somebody, at some point, since it was still sitting in the file a hundred years later. I photographed the three pages and opened Google's AI Studio where Gemini handles handwriting work the show has relied on all along. I want to flag something before I show you the result though. Everything this show has verified about Gemini's handwriting accuracy has been on English language documents, deeds, pension files, letters and cursive English. I could not find solid, independently tested numbers on how it performs on handwritten Chinese specifically, as opposed to typed or printed Chinese, which is a different and generally easier task. That gap is exactly why I treated this transcription as a first draft, not a verdict, and verified it harder than I would an English language document. I did not take its translation on faith without checking it. Before I trusted a single line of it, I ran a small check. I found two of the same Chinese characters recurring in different parts of the letter. I cross checked them myself against a Cantonese dictionary site and confirmed the translation was reading each one the same way every time. That's a small check. It's not the same as having a fluent reader review the whole thing. But it's the difference between using a tool and handing your judgment over to it. Here's exactly what I typed. Quote Here are three photographs of a handwritten letter in Chinese found inside this immigration case file. Please transcribe the original Chinese characters first, then give me an English translation. Note any names, place names, or family relationship terms exactly as they appear, since those details matter more here than a smooth translation would, end quote. The translation came back, and it wasn't from Eng anybody. It was signed by a woman identifying herself as his mother. She didn't call him Yuk Sing. She called him by a name that appeared nowhere else in that entire two hundred page file, Duke On. She named a village that didn't match the official file at all. She asked, in a few plain lines near the bottom of the second page, whether he was eating enough and whether the cold in America was as bad as people said. Ordinary, worried, mother to son questions, tucked inside a federal suspicion file that had never once asked him anything so kind. And she used a character, transliterated as Duke, that showed up nowhere in the coach to testimony. Perplexity had already told me to watch for exactly this, a shared generation name, the kind of single character that runs through the given names of an entire set of brothers and cousins in the same family. It's a naming pattern, not a coincidence, and it's one of the more reliable signals in Chinese genealogical research once you know to look for it. On its own, one letter wouldn't be enough to overturn a hundred years of official paperwork. But it wasn't on its own. It came paired with a generation name that fit the pattern and a village that matched nothing in the coach story. Together, that was enough. I stopped trusting the files official version. I started trusting the mothers. Here's where I have to tell you about the dead end, because I got this part wrong at first. My first assumption was simple. Aang, the paper name, must connect back to a real Aang family lineage somewhere. I figured the District Association Records perplexity has surfaced would confirm it, if I searched hard enough. So I spent real time chasing an Aang family genealogy that fit the file's official story. A household size that lined up with a coach testimony. It looked promising for about an hour. Then it fell apart. The Aang lineage I'd found had documented complete sons all accounted for elsewhere. There was no room in it for a boy named Yuk Singh at all. No matching lineage. No matching village. I had chased the family the file wanted me to believe in, not the family that had actually sent him. The pivot came from going back to that one line in the letter Jim and I had transcribed, the shared generation name. Once I searched using that detail instead, alongside the real village name the letter gave, the pieces started lining up. This is resolution of a conflicting evidence, the fourth element of the genealogical proof standard, right here in front of me. Two documents in the same file, telling two different stories, and the letter with the Was the one that held up. I want to slow down on one detail from that original nineteen twenty one transcript because it's the piece that has stayed with me since I first read it. During the interrogation, one of the inspectors asked seventeen year old Yuk Singh how many steps led up to the front door of the house he was claiming as his family home. Steps not the family's name, not the village, the number of steps up to a door, an ocean away that he had never once climbed, memorized from a coaching book because his real front door didn't matter to the men asking the questions. He answered anyway. He had to. Somewhere in that same building, at some point, his real front door had a real number of steps too. Nobody at Angel Island ever asked him about those. That's the whole shape of what a paper name costs. It isn't only a false answer on a government form. It's a true life, sitting one door over from the one anybody bothered to ask about. Before I move on, I want to be straight with you about something. Every piece of that transcript I've described came from the file itself. None of it came from anything I imagined about how he must have felt. That's the golden rule again, showing up in a different context than it did earlier. Claude built me a comparison table. Jim and I translated a letter. Perplexity pointed me toward the right kind of records. Not one of them told me what any of it meant. That reading and that verifying against the actual page images stayed mine the whole way through. So where did all of this land? We found the real family. The letters village, cross checked against a district association record perplexity helped me locate, matched a farming family with a surname Moi, four sons, sharing that same generation character. Duke On was the second son. Put the two pieces together, the given name from the letter and the surname from the record, and you get his actual name. Not Eng Yuk Singh, Moi Duke On. His own mother had written to him under that name in a letter that sat quietly inside the very file built to prove he was somebody else's child. Once I had the real surname and the real village, the rest of the record started behaving the way records are supposed to. A district association roster from the right decade listed a moy household of the right size in the right place. It's not a courtroom grade proof chain yet. There's more work here for his descendants to finish. But for the first time, the pieces were describing one family instead of two families stitched together by a broker's coaching book. We didn't get every answer. We don't have Duke On's exact birth date and the exact fate of the real Moi family back in China after he left is still an open thread, one his descendants are still pulling on. There's a whole second research project waiting in that thread alone. Did the real Moi family know where he ended up? Did letters keep arriving after this one? Or did the war and the decades in the distance eventually close that door too? The file doesn't say, and neither does anything else we found this time around. But we got the one thing that mattered most his real name. His real mother's words in her own hand after a hundred years of silence, and his real place in a family that had been waiting all along for someone to come looking. That's the honest shaped of an unexpected discovery. Not a clean bow on top of the mystery we opened with, but a truer, harder answer than the one the file wanted anyone to find. So here's what I want you to walk away with. If you've got a case where the paper trail feels unusually thick, and I mean genuinely unusual compared to everyone else in the family, don't assume that's bad luck. Ask why the government cared enough to build that file in the first place. And if part of that file is in a language you don't read, don't skip it. That's often where the real story is sitting, waiting for a tool that can finally open the door for you. That goes well beyond Chinese exclusion files, by the way. Anytime a government, a court, or an institution decided a group of people deserved extra scrutiny, it tends to leave behind an unusually thorough paper trail on the people it was watching. Recognizing that pattern is a skill that travels across record types, across countries, and across every heritage line this show covers. Homework for this week. Pull up one document in your own tree that never quite added up. A name, a birthplace, a date. Ask what would have to be true for that inconsistency to be the point rather than a mistake. You might not find a paper son. You might find nothing more than a tired clerk having a bad afternoon. Either way, you'll have practice the habit that matters. And for my Australian and UK listeners, this file type has a direct cousin in your own archives. Australia ran its own version of exclusion through the Immigration Restriction Act of nineteen oh one. Chinese residents needed a certificate exempting from dictation test or CEDT before they could leave the country and come home again. Those certificates, complete with photographs and fingerprints, are held today by the National Archives of Australia and they document the same kind of official suspicion we saw at Angel Island. For my UK listeners, the parallel runs through the Aliens Act of nineteen oh five and onward. Once Britain started registering immigrants formally, the home office built its own paper trail held today at the National Archives in Kew. The instinct to read it the same way applies there as much as it does in San Francisco. 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 use what we cover today, share this episode with them. That's the best way to help our community grow. If today's episode hit you somewhere, the companion guide goes a step further. I've built out the exact prompt sequence for spotting a coached interrogation versus a genuine one, along with a walkthrough for handling documents in a language you don't read yourself. The free episode stands on its own, but the companion guide is where the deeper toolkit lives. For everything you need, including every episode, our private Facebook community, companion guides, and the research lab, head over to ancestors andai dot 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.