AI for Genealogy: AI Signature Analysis Uncovers Identity Theft - Polish Immigrant Mystery Case Study

Discover how AI signature analysis solved a 120-year-old genealogy mystery involving identity theft, mining deaths, and desperate immigrants in Pennsylvania coal country. This complete tutorial demonstrates using free AI tools—Gemini, Claude, Perplexity, and ChatGPT—to analyze handwriting, compare documents, and uncover hidden family secrets.

THE MYSTERY: Two property deeds, same name, same land—but completely different signatures. One showed educated European penmanship, the other barely literate scrawl. Were these signed by the same person? AI revealed the shocking truth: identity substitution after a tragic mining death.

WHAT YOU'LL LEARN:

AI Signature Analysis: Use Gemini's free image analysis to objectively compare handwriting across documents. Get exact prompts for analyzing letter formation, pen pressure, and writing systems that distinguish one person from another.

Document Pattern Recognition: Learn Claude's powerful chronological analysis technique to spot identity transitions, witness changes, and gaps in the documentary record that reveal when someone assumed a new identity.

Historical Context Research: Master Perplexity's citation-backed research to understand why immigrants changed identities—mining accidents, unreported deaths, "paper sons," and substitute identities in early 1900s America.

Theory Generation: Use ChatGPT to systematically generate and evaluate multiple explanations for genealogical mysteries, avoiding tunnel vision and confirmation bias.

TECHNIQUES DEMONSTRATED: ✓ Free-tier AI tools only (Gemini, Claude, ChatGPT, Perplexity) ✓ Copy-paste ready prompts for immediate use ✓ Multi-tool workflow combining signature analysis, document comparison, and historical research ✓ Primary source verification methods ✓ Advanced genealogy proof standards with AI assistance

REAL GENEALOGY CASE STUDY: Follow the investigation of Jan Kowalski, Polish immigrant who purchased Pennsylvania farmland in 1897—then "sold" it in 1902 with a completely different signature. AI tools revealed coroner reports, mining accident records, and boarding house keeper documentation proving the original Jan died in 1901, and another immigrant assumed his identity to claim naturalization papers and property.

WHO THIS EPISODE IS FOR: → Genealogists with signature mysteries across multiple documents → Family historians comparing handwriting to determine if records match the same person → Researchers investigating immigrant identity changes, name variations, or missing relatives → Anyone stuck on brick walls where records don't align logically → Beginners wanting step-by-step AI genealogy workflows

HOMEWORK ASSIGNMENT: Compare signatures for one of your ancestors across multiple documents using the free Gemini analysis technique demonstrated in this episode.

KEYWORDS: AI genealogy, signature analysis, handwriting comparison, ChatGPT genealogy, Claude AI family history, Gemini genealogy tools, Perplexity research, immigrant identity theft, Polish genealogy, naturalization records analysis, genealogy AI tools, family history chatbots, AI document analysis, genealogy brick walls

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

There were two signatures. Two completely different signatures. Both claiming to be Jan Kowalski . I was staring at my computer screen at 2 o'clock in the morning, because that's when we genealogists do our best work, right? Comparing two legal documents I'd found in the Pennsylvania County archives. Both were land deeds. Both were dated within five years of each other. Both were signed by a man named Jan Kowalski  living in the exact same township. But they weren't signed by the same person. One signature, from an 1897 property deed, showed beautiful, educated European script. Confident loops. Perfect spacing. The handwriting of someone who'd learned penmanship in a proper school. The other, from a 1902 property cell, was barely literate. Shaky block letters. The K in Kowalski  looked like it was written by someone who wasn't quite sure how to form the letter. Now, I've been doing genealogy for quite some time. The K in K in K in I've seen plenty of handwriting deteriorate due to age, injury, or illness. But this wasn't deterioration. These were fundamentally different writing systems. Different hands. Different men. And here's what made my heart race. Both deeds involved the exact same parcel of land. The first deed showed Jan Kowalski  purchasing 40 acres outside Scranton, Pennsylvania in 1897. The second showed Jan Kowalski  selling that same 40 acres in 1902. According to the property records, it was the same man. Same name. Same location. Simple transaction. Bought land, sold land. Five years later. Except it wasn't the same man. It couldn't be. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host, Brian, and today we're unraveling one of the most disturbing mysteries I've encountered in my genealogy research. A mystery that reveals the dark side of immigrant experience in early 20th century America. A mystery where AI tools helped me connect dots I never would have seen, but where the truth still required old-fashioned detective work and historical context. My great-grandfather, Jan Kowalski , immigrated to Pennsylvania from Austrian Poland in 1892. He was 22, escaping poverty and conscription, seeking his fortune in America like thousands of young Polish men. I grew up hearing family stories about his hard work in the coal mines, his frugality, how he saved enough to buy a small farm. A classic immigrant success story. Except these stories started to unravel when I found these deeds. The 1897 deed, the purchase, showed a signature consistent with other documents I'd found from Jan's early years in America. His declaration of intention to naturalize, filed in 1895. A witness statement he'd made in a friend's naturalization case in 1896. All showed the same educated European handwriting. But that 1902 deed, the cell, had a signature that matched nothing else in my files. And when I started looking more carefully at documents after 1902, a disturbing pattern emerged. Jan's signatures after that date were either this barely literate scrawl or they disappeared entirely. Legal documents had witness marks instead of signatures. In 1897. Census records showed him as unable to read or write. But I had proof he could write. Beautifully. So, what happened between 1897 and 1902? 

Here's where this story gets fascinating. And where AI tools became absolutely crucial for spotting patterns I never would have caught on my own. because what I discovered wasn't just about handwriting. It was about identity theft in an era before social security numbers or photo IDs. It was about desperate immigrants and unscrupulous middlemen. It was about a practice so common in early 1900s immigrant communities that historians have a name for it. Paper sons and substitute identities. Before we dive into the AI tools I used, let me establish something critical. AI is your research assistant, not your researcher. That's our golden rule here. What I'm about to show you is how AI helped me analyze patterns, spot inconsistencies, and generate research theories. But the actual verification came from primary sources, Historical context and good old fashioned genealogical proof standards. The techniques I'm going to demonstrate work for any situation where you're comparing signatures across documents, trying to determine if you're actually following the same person, or piecing together while records don't align the way they should. So here's what I did: The first challenge was confirming what my gut was screaming. These signatures were from two different people. I needed objective analysis before I went down what might be a rabbit hole. For that, I turned to Gemini, Google's free AI tool that has impressive image analysis capabilities. Let me show you exactly how I used it. Ok, so I had high resolution scans of both signatures from the county archives, the 1897 deed and the 1902 deed. I'd photograph them carefully under good light, making sure the pen strokes were clearly visible. Now, I could have just trusted my years of experience reading old documents and declared these were different people. But that's not how good genealogy works. We need evidence. Multiple pieces of evidence. And we need to rule out other explanations first. Could the 1902 signature be

from the same man after an injury? stroke? Could advancing age have degraded his handwriting that severely in just five years? From age 32 to 37? Unlikely but possible. I needed systematic analysis. I opened Gemini's free interface and use this exact prompt. And I want you to pay attention to how I structured it. Quote I'm uploading two signature images from Pennsylvania property deeds. The first is dated 1897, signed by Jan Kowalski , age 32. The second is dated 1902, supposedly signed by the same Jan Kowalski , now age 37. Please analyze these signatures for: One, fundamental differences in letter formation and writing style. Two, whether these could be from the same person, accounting for age or physical changes. Three, specific features that distinguish the two hands. Four, your professional assessment of whether these are from the same writer or different writers. End quote. See what I did there? I gave Gemini's? I gave Gemini the historical context. Dates, ages, document types. I asked it to consider alternative explanations like aging or injury. And I requested a specific actionable analysis, not just a yes or no answer. Gemini's response was thorough, and damning. It identified what it called incompatible fundamental characteristics. 

The 1907 signs werewolves. The 1907 signature showed what handwriting experts called copybook style. Evidence of formal European education. Connected letters. Consistent baseline. Controlled pin pressure. The capital J and K were formed using traditional Polish school methods from the 1880s. The 1907 signs. And for geometry. These signs of driving the 1910s. And the 1910s. And the 1910s. And the 1910s. Uh, the 1902 signature showed emergent literacy characteristics. Someone who'd learned to write as an adult, probably by copying, disconnected block letters, wavering baseline, uneven pressure suggesting uncertainty. The capital letters were formed using what looked like American methods, but executed clumsily. Gemini's conclusion. Gemini's conclusion. Quote, "These signatures show fundamentally different educational backgrounds in motor skill development. The probability they were written by the same person, even accounting for injury or illness, is extremely low." End quote. Now, I had objective confirmation, but I needed to organize every document I had about Jan Kowalski , chronologically, and look for other changes around 1897 to 1902. This is where Claude became invaluable. Claude excels at taking scattered information and creating structured analysis. I uploaded images of 12 different documents spanning 1892 to 1925. Naturalizations. Deeds. Court testimonies. Affidavits. Anything with a signature or mark. Then I asked Claude to build me a comprehensive comparison. Here's the prompt I used. Quote, "I'm researching Jan Kowalski , Polish immigrant to Pennsylvania, the various immigrant to Pennsylvania, arrived 1892. I'm uploading 12 documents from 1892 to 1925. Create a detailed chronological table with these columns: Date, Document, Type, Signature Type, Handwriting Style, Name Spelling, Other Witnesses Listed, and Notable Changes. Flag any transitions or patterns that suggest identity changes or documentation issues," end quote. And friends, what Claude revealed was stunning. Let me walk you through the pattern. 1895 to 1897, four All signed Jan Kowalski in educated European script. Witnesses were other Polish immigrants, mostly coal miners. 1898 to 1901, no documents found. This gap would become significant. 1902, property sale. Signature barely literate. Witnesses were English-speaking Americans, not Polish immigrants. 1903 to 1909, six documents. All showed either barely literate signatures or witness marks. Jan making an X instead of signing. Witnesses increasingly non-Polish. 1910 to 1925, eight documents. Not a single full signature. All marked with X or signed by witnesses. Census records listed him as unable to read or write. Do you see it? The educated Jan Kowalski, who could write beautifully, vanished between 1897 and 1902. After 1902, the man living under that name was functionally illiterate. But, here's what made my blood run cold. That property sale in 1902 wasn't just suspicious because of the signature. When I looked carefully at the deed itself, the sale price was remarkably low. Jan had purchased 40 acres for $800 in 1897, a fair price for farmland in that area. He sold it for $300 in 1902. Nobody sells land at a 62% loss after five years unless they're desperate, or unless they're not actually the person who bought it. I needed to understand what was happening in Pennsylvania immigrant communities during this period. What would cause someone to assume another person's identity? What legal or economic forces were at play? This is where perplexity became my research assistant. Its ability to search historical sources with citations meant I could verify everything I found. Let me show you what I discovered about the darker side of immigrant life in 1900s Pennsylvania. 

I opened perplexity. Remember, the free version gives you five searches per day, which is usually plenty for focused research. And I started asking very specific questions about Polish immigrant experiences around the turn of the century. My first search was targeted, quote, Pennsylvania coal mining accidents, Polish immigrants, 1898 to 1902 death rates, end quote. What came back was heartbreaking and illuminating. Between 1897 and 1902, Pennsylvania's anthracite coal region experienced what one contemporary report called an epidemic of industrial death. Mine safety regulations were virtually non-existent. The accident rate for foreign born miners, especially Poles, Slavics, and Italians, were significantly higher than for native born workers because they were assigned the most dangerous positions underground. But here's what caught my attention. A 1901 report from the Pennsylvania Bureau of Industrial Statistics noted that many mine deaths of immigrant workers went unreported or were misreported because the victims carried no identification papers and left no official records. Some had been using assumed names to avoid military conscription back in Europe. Some had used false identities to get into the country. And here's the devastating part. When an unidentified immigrant died in a mine accident, the mining company often didn't bother investigating. They had buried the body in a pauper's grave and moved on. I needed to know more about identity substitution. My next perplexity search, quote, Polish immigrants, identity theft, Pennsylvania, 1900s, substitute names, dead miners, end quote. What I found described a practice that was apparently widespread, though rarely documented in official records. When an immigrant died without family in America, others sometimes assumed their identity, especially if the deceased had valuable documents or property. One historical article cited cases of immigrants taking over the identities of dead countrymen to claim their savings accounts, their property claims, even their naturalization papers. In immigrant communities where record keeping was informal and photo identification didn't exist, this was shockingly easy to do. I was starting to form a hypothesis, but I needed to test it systematically. This is where ChatGPT became my thinking partner. I asked ChatGPT to help me generate and evaluate theories based on the evidence. Quote, Based on these documented facts, Polish immigrant Jan Kowalski shows educated signature 1895 to 1897, purchases property 1897, then disappears from records 1898 to 1901, reappears 1902 with barely literate signature, and sells same property at huge loss. All subsequent documents show illiteracy. Historical Context High death rates among Polish coal miners, identity substitution, documented in immigrant communities. Generate five possible explanations ranked by probability. End quote. ChatGPT returned five scenarios, and I want to walk you through them because this kind of systematic theory generation is where AI really earns its keep. Theory 1, highest probability Original Jan Kowalski died between 1897 to 1902, possibly in a mining accident. Another Polish immigrant, perhaps with a similar name or from the same region, assumed his identity to claim the land, then sold it quickly to extract value before disappearing. The signature difference reflects the different individuals. Theory 2 Original Jan Kowalski suffered catastrophic injury between 1897 to 1902, affecting his ability to write. Sold land at loss due to disability. Subsequent illiteracy due to injury-related cognitive impairment. However, this doesn't explain why his entire education apparently vanished. Theory 3 Original Jan deliberately gave his identity to another immigrant, possibly for payment or to help a desperate compatriot, and left the area. The substitute lived as Jan Kowalski thereafter. This was documented in some immigrant communities as a form of charity or commercial transaction. 4 Original Jan was involved in legal trouble and fled leaving another immigrant to deal with his affairs and assume his identity. The land sale was part of liquidating his assets. Theory 5 Lowest probability The 1897 signature is actually forged by a clerk or witness, and the 1902 signature represents Jan's actual writing ability. This doesn't align with multiple documents from 1895 to 1897 showing consistent, educated handwriting. Theory 1 kept nagging at me. Death in the mines. It fit the timeline. It explained everything. But I needed proof. I went back to Claude and asked it to analyze something I'd overlooked. The witnesses on these documents. Quote, Review all 12 documents and create a social network analysis of witnesses. Who witnessed which documents? Are there patterns in witness relationships over time? Do any witnesses appear on both pre-1902 and post-1902 documents? End quote. Claude's analysis revealed something crucial. Pre-1902 documents, the educated Jan, were consistently witnessed by men with Polish surnames. These were clearly Jan's social circle in the immigrant community. Post-1902 documents, the illiterate Jan, were witnessed almost exclusively by English-speaking Americans. William Morrison, James Sullivan, Thomas McKenna. A completely different social network. Except for one name that bridged both periods, Stanislav Boric. Boric witnessed Jan's 1896 court testimony, educated signature, and Boric witnessed the 1902 property sale, illiterate signature. He was present at both identities. Who was Stanislav Boric? And what did he know about the two Jan Kalowskis? I asked Perplexity to help me find information about him. Quote, Stanislav Boric, Polish immigrant, Pennsylvania coal mines, 1890s to 1900s. End quote. And here's where this gets interesting. Perplexity found a reference to a Stanislav Boric in a 1903 Pennsylvania death registry. Not for Boric himself, but for a death he reported to authorities. He was listed as a boarding house keeper in Scranton. Boarding house keeper. In immigrant communities, boarding house keepers were central figures. They housed newly arrived immigrants, helped them find work, handled money transfers back to Europe, sometimes served as informal bankers and notaries. They knew everyone in the community. If Boric ran a boarding house, he would have known the original Jan Kalowski. He would have known if Jan died. And he would have been perfectly positioned to facilitate an identity transfer to another immigrant. But I needed to verify this theory. I needed mining accident reports. I needed death certificates. I needed proof that the original Jan Kowalski died between 1897 and 1902. I asked Perplexity, quote, Where would Pennsylvania mining accident records 1897 to 1902 be archived? Specifically, Scranton Anthracite region, end quote. Perplexity directed me to three repositories, the Pennsylvania State Archives, the Anthracite Heritage Museum, and, crucially, the records of the United Mine Workers of America, which had begun keeping their own accident reports in the late 1890s because company reports were so unreliable. And that's when I hit the archives. Not digitally, physically. Because some records, especially from this era, haven't been digitized. I contacted the Pennsylvania State Archives in Harrisburg with a specific mission, find evidence that Jan Kowalski died in a mining accident between 1897 and 1902. And friends, what I found in those dusty ledgers? It changed everything. 

The Pennsylvania State Archives holds mining accident reports that were filed by local coroners, not by mining companies. This is crucial because company reports often minimize deaths or fail to report them at all. But coroners, by law, had to investigate any suspicious deaths. I requested all coroner reports from Lackawanna County, where Scranton is located, from 1898 to 1902. What I was looking for was any Jan Kowalski, any Kowalski at all, who died in that period. And I found him. June 14th, 1901. Coroner's Report. John Kowalski, approximately 31 years old, Polish nationality, killed by Rockfall in Delaware in Hudson Coal Company Mine Shaft No. 7. The report was brief. They all were in that era. But it included crucial details. The deceased was identified by Stanislaw Boric, boarding housekeeper, who stated the deceased had resided at his establishment for approximately six years. Six years from 1901 takes you back to 1895, exactly when my Jan Kowalski would have arrived in the area after filing his declaration of intention. The coroner's report noted, Deceased had no family in America. Personal effects limited. Body claimed by Polish Mutual Aid Society for Burial. No family. Personal effects limited. This was my educated Jan Kowalski, the man who could write beautifully, who had saved enough to buy 40 acres of farmland, who had his whole American dream ahead of him. And he was crushed by falling rock in a coal mine at age 31. But here's what the coroner's report didn't mention. What happened to Jan's property? His 40 acres? His documented claim to American residents? I went back to the land records and looked more carefully at the 1902 sale, the one signed with barely literate handwriting. The transaction was handled by Stanislaw Boric as a representative of the deceased owner's estate. Except Jan wasn't listed as deceased in the land records. The sale was recorded as if Jan Kowalski himself was selling the property. Boric was just a facilitator, the witness. Here's what I think happened, piecing together fragments from historical practice and documented cases. When Jan Kowalski died in June 1901, he left property but no family. Stanislaw Boric, as the boarding housekeeper and community leader, was responsible for settling his affairs. The property needed to be sold. It couldn't just sit there generating taxes. At the same time, there was another Polish immigrant, perhaps newly arrived, perhaps desperate, perhaps running from something in Europe, who needed an identity. Someone who could never get naturalization papers under his own name. Someone who needed to become American but couldn't do it legally. Boric saw an opportunity. Or maybe it was common practice. Jan Kowalski was dead, but his identity was still alive on paper. His declaration of intention was on file. He had property in his name. He was, on paper, an established resident. So, Boric arranged for this other immigrant to assume the identity. The man learned to write Jan Kowalski as best he could, but he'd never had formal education like the original Jan. He could form the letters, but not with the same fluency. They sold the land quickly, at a loss, because speed mattered more than price. And the substitute Jan Kowalski faded into the immigrant community. Living out the rest of his life under a dead man's name. the 1910 census, there's a Jan Kowalski, age 45, living in Scranton, working in the coal mines, listed as unable to read or write English or Polish. That's the substitute. Living the life of a man who died in 1901. The real Jan Kowalski, the educated one, is buried somewhere in a mass grave at a Polish Catholic cemetery his headstone probably long gone, his identity stolen by necessity and desperation. When I traced my family line forward, I realized I'm not actually descended from the man I thought was my great-grandfather. The woman my family always said was Jan's daughter, my grandmother, was actually the daughter of the substitute Jan, a different man entirely. My real great-grandfather died childless at 31. My actual great-grandfather is the illiterate coal miner who lived the rest of his life under a dead man's name. you know what? I understand why it happened. These were desperate men in a desperate situation. Identity theft sounds criminal to us now, but in 1901, in Pennsylvania Cole County, it was sometimes an act of survival. The substitute Jan needed papers. The dead Jan's identity would have gone to waste. Boric facilitated what might have seemed like a practical solution. But it makes tracing my family history infinitely more complicated. Because every record before 1902 is about a different man than every record after 1902. And this is exactly why signature analysis matters. Why comparing documents matters. Why AI tools that can help you spot patterns matters. Because sometimes the signatures reveal secrets that census records and family stories never will. Let me show you how you can apply these same techniques to your own brick walls. 

Okay, let's break down the investigative techniques we use today so you can apply them to your own genealogy mysteries. First, we use Gemini for objective signature analysis. When you suspect you're looking at different people using the same name, Upload clear images and ask for a systematic comparison. Gemini can spot differences in letter formations, pen pressure, and writing systems that might escape casual observation. Second, we use Claude to create chronological pattern analysis across multiple documents. This is invaluable when you're trying to spot transitions, gaps, or changes in your ancestor's documented life. Upload everything you have. Ask for structure comparison and let AI reveal patterns you might miss. Third, we use perplexity to research historical context. The social, economic, and legal forces that might explain puzzling behavior. AI is your research assistant, not your researcher. But it can point you toward historical sources that explain why immigrants took desperate measures. Fourth, we use ChatGPT 

When you're stuck, having a thinking partner that can consider multiple scenarios simultaneously helps you avoid confirmation bias and tunnel vision. But here's what AI didn't do. AI didn't contact Harrisburg and get those coroner reports. AI didn't cross-reference land deeds with death records. AI didn't verify the timeline against mining accident statistics. That was old-fashioned genealogy legwork. 

The power comes from combining AI's pattern recognition capabilities with traditional research methods and primary source verification. Here's your homework. Pull out legal documents for one of your ancestors, deeds, wills, court testimonies, affidavits, anything they personally signed. Compare signatures across time. Do they stay consistent? Do they change? Are there gaps in the documentary record that can side with signature changes? Upload signature images to Gemini and ask for analysis. You might be surprised by what you find. Sometimes signature changes reveal injury or aging. Sometimes they reveal literacy development. And sometimes, like in my case, they reveal that you're actually following two different people. If you discover something puzzling, share it in our Facebook group, Ancestors and Algorithms AI for Genealogy. Or email me at ancestorsandai at gmail.com. Your mystery might become a future episode. Thank you for listening to Ancestors and Algorithms. If this episode helped you think differently about your research, please share it with a fellow genealogist and leave a review on your podcast platform. And remember, AI is your research assistant, not your researcher. Use it to spot patterns and generate theories, but always verify with primary sources and real historical context. I'm your host, Brian, and I'll see you next week for another journey into the past, powered by the future. Until next time, happy researching.