So, I'm staring at a 47-page probate file from 1892. 47 pages of cramped handwriting, legal jargon, inventories of farm equipment, and names I'd never seen before, scattered across every page. My great-great-grandfather's estate file. I'd have this document for three years. Three years. And every time I opened it, I'd read a few pages, get confused by the legal language, make a few notes, and close it again. I knew there were secrets buried in those pages. Names of relatives I've never found. Clues about property disputes. Maybe even the answer to why half the family stopped speaking to the other half. But 47 pages? Then I discovered an AI tool that changed everything. Not because it did my research for me. Remember, AI is your research assistant, not your researcher. But because it read those 47 pages and became an instant expert on my family's 1892 legal drama. 20 minutes later, I had answers to questions I didn't even know to ask. That tool is Google Notebook LM. And today, I'm going to show you exactly how it works. Welcome to Ancestors and Algorithms, where family history meets artificial intelligence. I'm your host, Brian. And today, we're diving deep into what I believe is one of the most underrated and powerful AI tools available to genealogists right now. If you've heard of Notebook LM, you might know it as that tool that makes podcasts from your documents. And yes, it does that, and we'll get there. But that's like saying a smartphone is just for making phone calls. There's so much more happening here. Here's what makes Notebook LM fundamentally different from ChatGPT, CLOD, Gemini, or any other AI tool you've used. Those tools answer questions from their training data. Millions of books, websites, and documents they've read. Which means they can make things up. They can hallucinate ancestors who never existed. They can confidently tell you your great-grandmother immigrated in 1892 when she actually arrived in 1912. Notebook LM doesn't work that way. Notebook LM only knows what you give it. You upload your documents, your census transcriptions, your probate files, your research notes, and it becomes an expert on those documents. Every answer comes with a citation pointing back to your sources. It can't make things up because it only has your research to draw from. For genealogists who care about proof and verification, that's revolutionary. Today, I'm going to walk you through everything. What Notebook LM is, how to set it up, the features that matter most for genealogy, and the advanced techniques that will transform how you work with your research. I've got a mystery to solve along the way. And by the end of this episode, you'll have everything you need to try this yourself. Let's start at the beginning. 

Google Notebook LM launched as an experimental tool in 2025, and it's evolved dramatically since then. As of January 2026, it runs on Gemini 3, Google's most advanced AI model, and it's become a full-fledged research companion. The LM stands for Language Model, but think of it as your personal research librarian. You give it your documents, and it reads everything, understands the connections, and stands ready to answer any question about those specific sources. When you ask ChatGPT about your ancestor, John Smith, it searches its vast training data and tries to give you a helpful answer. It might tell you about naming patterns, suggest record types to search, or explain historical context. That's valuable. But it doesn't know your John Smith. When you ask Notebook LM about John Smith, after uploading your census records, your church records, your family notes, it tells you specifically what your documents say about your John Smith. And it shows you exactly which document that information came from. See the difference? One is a knowledgeable stranger making educated guesses. The other is an assistant who has actually read all your files. So, what can you upload? Notebook LM accepts almost everything you'd want to use for genealogy research. PDF documents, which means census transcriptions, downloaded records, research reports, anything you've saved as a PDF. Google Docs and Google Slides. So, your ongoing research notes can go right in. Website URLs. Found a great article about migration patterns in your ancestors' region? Paste the link and Notebook LM reads it. YouTube videos. It pulls the transcript automatically. This is incredible for genealogy lectures and tutorials. Google Sheets. Added just last month. You can now upload spreadsheet data. Microsoft Word Documents, also new as of late 2025. And here's one that surprised me. Images. You can upload photographs of documents, and Notebook LM will analyze them. The transcription isn't perfect for difficult handwriting. I'd still use Gemini and Google AI Studio for serious transcription work. But for printed documents or clear images, it works remarkably well. Now, I know some of you are already thinking, what's this going to cost me? Here's the good news. Notebook LM's free tier is genuinely generous for most genealogists. Say that five times fast. On the free version, you can create up to 100 notebooks. Each notebook can hold 50 sources. Each source can contain up to 500,000 words. You get 50 chat queries per day and three audio generations per day. me put that into perspective. A typical census transcription might be 500 words. A detailed research report might be 5,000 words. A lengthy probate file might be 20,000 words. You could upload an enormous amount of family research before hitting any limits. There is a paid tier, Notebook LM Plus, which comes bundled with Google One AI Premium for $19.99 a month. It increases your limits to 500 notebooks, 300 sources per notebook, 500 daily queries, and 20 audio generations per day. But honestly, start with the free tier. Most hobbyist genealogists will never need more. Setting up is simple. Go to Notebooklm. google.com and sign in with your Google account. That's it. No credit card, no trial period, no complicated setup. Click New Notebook and give it a name. I recommend organizing notebooks by family line or research question, not by document type. So, Smith Family Hyphen Ohio 1850-1900 rather than census records. Then start adding sources. Click the plus sign, upload your PDF, paste your URLs, connect your Google Docs. The more relevant sources you add, the smarter your notebook becomes. Within minutes, you're ready to start asking questions. 

Let me take you back to that 47-page probate file I mentioned at the beginning. My great-great-grandfather, William, died in 1892 in rural Indiana. I'd found his probate file on FamilySearch years ago. It was digitized, which was a blessing, but the sheer volume of pages intimidated me. Every time I tried to work through it, I'd get lost in the legal language. Whereas, the errors of the first part. Pursuant to the statutes of the state of Indiana. My eyes would glaze over, but there was something in that file that haunted me. Family lore said that after William died, two of his sons never spoke again. Something happened during the settling of the estate that tore the family apart. My grandmother remembered being told never to mention Uncle Henry around her father. When I asked why, she just shook her head and said, The farm. It was about the farm. That's all I had. The farm. I knew the answer was somewhere in those 47 pages. I just couldn't find it. So, I created a new notebook in Notebook LM. I called it William Anderson Estate 1892. First, I uploaded the complete probate file as a PDF. Then I added a few supporting documents. The 1880 census showing William's household, an 1890 county history that mentioned the family, and my own research notes about the children and their families. Four sources. That's all it took. Within about 30 seconds, Notebook LM had processed everything. And here's where it started to get interesting. Before I even asked a question, Notebook LM automatically generated a summary of my sources. Just a paragraph at the top of the screen. And in that summary, it mentions something I'd completely missed in three years of having this document. Quote, The estate documents reveal a contested claim between heirs regarding the disposition of the home farm. Henry Anderson filed a formal objection to the administrator's proposed distribution on March 15, 1893, citing a verbal agreement with the deceased that predated the written will. A verbal agreement that predated the will. I'd read this file multiple times. How had I missed that? But when you're reading 47 pages of legal jargon, looking for specific names and dates, it's easy to miss the significance of procedural documents buried in the middle. Notebook LM had read everything and understood the narrative. Thank you. Now, I started asking questions, and this is where Notebook LM becomes genuinely powerful for genealogy research. I typed, quote, what was Henry Anderson's objection to the estate distribution, and what evidence did he present, end quote. The answer came back with citations to three specific pages in the probate file. It explained that Henry claimed his father had promised him the home farm in exchange for staying and working the land while his brothers left for the city. He provided the names of two witnesses who would testify to this verbal agreement. Then I asked, quote, who were the witnesses Henry named, and what was their relationship to the family, end quote. Notebook LM pulled from both the probate file and the 1880 census I'd uploaded. One witness was a neighboring farmer. The other, and this is where I felt the hair stand up on my arms, was William's own sister. Henry had his aunt ready to testify that her brother had promised him the farm. The family drama was right there in the documents. I just hadn't connected the pieces. I asked question after question, and Notebook LM traced the entire story through those 47 pages. How was the objection resolved? The state went to a hearing. The court ultimately sided with the written will over the verbal agreement. The farm was sold and the proceeds divided equally among all heirs. Which heirs received payments and what amounts? Notebook LM generated a table showing each heir's name and their share. Henry received $847. 32, exactly the same as his brother's. The farm he'd worked for 20 years, believing it would be his, was sold to a stranger. Did any heirs refuse their payment or file additional objections? And here was the answer to the family mystery. Quote, the documents show that Henry Anderson's signature is absent from the final receipt of disbursement dated November 1893. A notation indicates the payment was delivered via the county sheriff. End quote. Henry was so angry he wouldn't even sign for his share of the money. They had to send the sheriff to deliver it to him. No wonder he never spoke to his brothers again. Now, I want to pause here because this is crucial. I've just learned an incredible family story. But remember our golden rule. AI is your research assistant, not your researcher. Everything Notebook LM told me came with citations. It told me exactly which page each piece of information came from. So, I went back to the original images on family search and verified every single claim. The verbal agreement objection? Page 23. The witness names? Page 24. The final disbursement with Henry's missing signature in the sheriff's notation? Page 41. It was all there. Notebook LM hadn't invented anything. It had simply read the documents more carefully than I had and connected the dots. That's the difference between an AI that helps you research and an AI that does research for you. Notebook LM helped me understand documents I already had. The verification was mine to do. 

Asking questions is just the beginning. the right side of the Notebook LM interface, you'll find something called the studio panel. This is where the real magic happens for genealogists. Let me walk you through the features that have transformed my research. Feature number one, timeline generation. You can ask Notebook LM to create a timeline from your sources. For genealogy, this is extraordinary. Upload census records from 1850, 1860, 1870, and 1880 for the same family. Add a death certificate, a marriage record, and an obituary. Then ask Notebook LM to create a chronological timeline of all events. It will pull every date from every document and arrange them in order. Suddenly, you can see the shape of an ancestor's life laid out clearly. And you can spot gaps. Wait, there's nothing between 1862 and 1867. Where were they during those five years? The timeline becomes a research tool, showing you exactly where to focus next. Feature number two is mind maps. This feature creates a visual representation of the connections in your sources. Click Mind Map in the studio panel, and Notebook LM generates an interactive diagram. For genealogy, I use this to visualize family relationships. Upload several family group sheets or research notes, and the mind map shows you how everyone connects. Click on a node, and it expands to show you what your sources say about that person. I've found unknown connections this way. When you see the visual map, patterns emerge that aren't obvious when you're reading documents one at a time. Feature number three, data tables. This feature was added just last month in December 2025, and it's perfect for comparing genealogical records. Upload census records from multiple years, then ask Notebook-LM to create a comparison table. Specify the columns. Name, age, birthplace, occupation, state of residence. Notebook-LM extracts the information from each document and arranges it in a structured table. Now, you can see at a glance that your ancestor's reported age doesn't add up consistently, or that their birthplace changed from Pennsylvania to Virginia to Maryland across three censuses. Those inconsistencies are research gold. They tell you something's going on that needs investigation. Even better, you can export these tables directly to Google Sheets for further analysis. Feature number four, deep research. This is one of the newer features, and it's different from everything else. Deep research doesn't just analyze your uploaded sources. It goes out to the web to find related information. Here's how I use it for genealogy. I'll upload my research on a particular family, then ask deep research to find historical context about their community. Quote, research the economic conditions in Shelby County, Indiana in the 1880s. What industries were present? What events affected farmers during this period, end quote. Deep research generates a report with citations to dozens of web sources, newspaper articles, historical society pages, academic articles. It's like having a research assistant who goes to the library for you. The key is that you're not asking it to research your family. That's your job. You're asking it to research the world your family lived in, the context of the world you're having. Which is something that's going to ask you about. 

Okay, let's talk about that feature that made Notebook LM famous, the audio overview. Click audio overview in the studio panel and Notebook LM generates a podcast style conversation about your sources. Two AI voices, one male, one female discuss the key points, make observations, express surprise at interesting details, and summarize everything in an engaging format. Now, you might be thinking, why would I want a podcast about my research? Three reasons. First, learning. When you hear information presented conversationally, you process it differently than when you read it. I've caught details and audio overviews that I missed while reading the same documents. Second, sharing. Have a family member who's interested in the family history but won't read the documents? Generate an audio overview and send it to them. It's eight minutes of engaging conversation about your research. Suddenly, they're hooked. Third, review. I generate audio overviews before research sessions as a refresher. I can listen while driving, while doing dishes, while walking. It keeps the research fresh in my mind. You can customize the overview too. Tell it to focus on specific topics. Adjust the tone. Make it brief or detailed. And you can download the result as an audio file. Feature number six, infographics and slide decks. These features were added in November 2025 and the genealogy community has embraced them enthusiastically. Upload a application file and ask Notebook LM to create an infographic summarizing the military service. It generates a visual representation, a single image with the key dates, locations, units, and service details arranged attractively. These infographics are perfect for sharing on social media, including in family histories or presenting at genealogy society meetings. You're not spending hours in Canva trying make something look professional. Notebook LM does it in seconds. The slide deck feature is similar, but features a full presentation. Upload your research on a family name and Notebook LM generates 12 to 15 slides telling the story. It includes the facts from your documents, adds relevant imagery, and creates a narrative flow. I'm not saying these replace careful, human-crafted presentations, but as a starting point. They're remarkable. And lastly, feature number seven, flashcards and quizzes. This might seem like an odd feature for genealogy, but hear me out. Generate flashcards from your research, and you've created a study tool for your own family history. Who married whom? What years they immigrate? Where did they settle? If you're preparing for a research trip or getting ready to interview an elderly relative, flashcards help you internalize the details so you can focus on the conversation, not on checking your notes. 

This next segment includes some advanced techniques and prompts. Technique number one: the comparison query. Here's a prompt structure that's become invaluable in my research. "Compare all documents related to, ancestor name, and identify any inconsistency in dates, ages, names, birthplace or family relationships. Present the inconsistencies in a table format with the source of each conflicting piece of information." End quote. NotebookLM excels at this kind of cross-document analysis. It's doing in seconds what would take you hours of careful comparison. Technique number two, the gap analysis. After you've added all your sources for a particular ancestor, try this prompt. Quote, based on the documents provided, create a timeline of, ancestors' names, documented life events. Then identify time periods with no documentation and suggest what types of records might fill those gaps, end quote. This turns NotebookLM into a research planning tool. It shows you exactly where your documentation is thin and points you toward solutions. Technique number three, the relationship mapper. Quote, identify all individuals mentioned in these documents and their stated or implied relationships to ancestor name. Include witnesses, neighbors, and business associates, not just family members, end quote. Genealogists know that the fan club, friends, associates, and neighbors, often holds the key to breaking brick walls. This prompt helps you identify your ancestor's social network from your existing documents. Technique number four, context loading. This is an advanced technique that dramatically improves NotebookLM's analysis. Before uploading your genealogical documents, also upload reference materials. A guide to reading census reports. An explanation of probate terminology. Historical context about the time period and location. When NotebookLM has that background knowledge available, it interprets your documents more accurately. It understands that consumption on a death certificate means tuberculosis. It knows that laborer in 1880 had different implications than laborer in 1920. And finally, technique number five, the narrative generator. Quote, using only the facts documented in these sources, write a 500-word biographical narrative of, and sister name, in chronological order. Cite each fact with its source document, end quote. This doesn't replace your own writing, but it gives you a structured draft that you can refine. Every claim is sourced, so you know exactly what you're working with. Here's what NotebookLM cannot do. Let me be clear about the limitations. NotebookLM cannot transcribe difficult handwriting with high accuracy. For that, I still recommend using Gemini in Google AI Studio, which now achieves near-human-level transcription quality. NotebookLM cannot access external databases. It doesn't search ancestry or family search. It only knows what you give it. NotebookLM cannot verify information against records you haven't provided. If you only upload transcriptions, it can't check them against original images. And NotebookLM can make mistakes in interpretation. I've seen it occasionally misidentify relationships or confuse people with similar names. This is why verification remains essential. Always check citations. Always confirm with original sources. The golden rule applies here as much as anywhere. AI is your research assistant, not your researcher. 

Here's what I want you to do after this episode. Go to NotebookLM. google.com Sign in with your Google account. Create a notebook named for one ancestor or one family line. Upload three to five documents. Maybe census records from different years. Maybe a probate file you've been meaning to dig into. Maybe your own research notes that you've compiled over time. Once the documents are processed, start with this question. Quote, summarize what these documents reveal about ancestor name and their family. End quote. Read the summary. Check the citations. Then ask follow-up questions based on what surprises you or what you want to know more about. That's it. That's your starting point. For this week's challenge, I want you to take one complex document. A probate file. A pension application. A lengthy land record. Something you've avoided because it seemed overwhelming. Upload it to NotebookLM and ask, Quote, what are the three most significant genealogical revelations in this document? End quote. Then verify those revelations against the original. Use the citations NotebookLM provides to check every claim. Share what you discovered in our Facebook group, Ancestors and Algorithms, AI for Genealogy. Use the hashtag, notebook. lm breakthrough. I want to hear what you're finding. want to leave you with something that struck me as I prepared this episode. For years, genealogists have talked about the problem of accumulated research. We gather documents, download images, take notes, and then we can't keep track of it all. The knowledge is scattered across files and folders and websites and our own memories. NotebookLM offers a different model. Upload your research and it becomes a queryable database. Ask any question. Get answers with citations. Generate summaries and timelines and visualizations. Your research stops being a pile of documents and starts being a living, interactive knowledge base. That's not replacing the human genealogist. That's amplifying what the human genealogist can do. Thank you so much for listening to Ancestors and Algorithms. If you found this episode helpful, please leave a review wherever you listen to podcasts. It truly helps other family historians find us. Don't forget to join our Facebook group. Ancestors and Algorithms, AI for Genealogy. Where you can share your Notebook LM discoveries, ask questions and connect with other researchers exploring these tools. We're over 1,000 members now and the conversations are incredible. Remember, AI is your research assistant, not your researcher. Notebook LM doesn't do your genealogy for you. It helps you understand documents you already have, find connections you might have missed, and organize your research in ways that weren't possible before. The verification, the analysis, the conclusions, those remain yours. Oh, and that 47-page probate file? I finally wrote to my cousin, Henry's great-great-grandson, and shared what I found. Turns out, his side of the family had always told a different version of the story. They said Henry was cheated. The written will, they claimed, had been influenced by the city brothers who came back just before William died. Was there truth to that? I don't know yet. But now I have a new mystery to solve. New documents to find. New questions to ask. That's genealogy, isn't it? Every answer opens three more questions. I'm your host, Brian. I'll see you next week for another journey into the past, powered by the future. Until then, happy researching.