The Case File: Four Generations of Descendants of Anne Lamb and Her Two Husbands

My ancestor Anne Lamb married twice. Her first husband was George Elliott (1732); her second, Henry Dawson (1745). Between the two marriages, she left behind a huge number of descendants — spread across Yorkshire parishes, and eventually, if the pattern holds, much further afield.

I’ve been researching this family for years. I have 702 individuals exported from my family tree as Anne’s descendants (including spouses). Some are well-documented. Many have gaps.

Is it necessary to trace four generations from my six-times-great-grandparents? Strictly speaking, no. But for piecing together DNA matches, having that depth of tree makes the job significantly easier. And do I want to? Absolutely. It's been a privilege and a pleasure to spend time in their part of the world.

The Central Mystery: Where Did They All End Up?

Anne Lamb had 12 surviving children at the time of her death in 1797, most of them leaving large numbers of descendants. I've written separately about what happened after the 1810 epidemic, which took her husband, son, and daughter-in-law: Six orphans, three guardians — but who actually raised them?

According to NotebookLM’s first pass, 58 of her 88 grandchildren have no death information recorded. I still don’t know where many of her great-grandchildren ended up. That’s the mystery — and NotebookLM is already helping me see the visible gaps.

The Case Lead: Lisa Rex

As the research lead, I keep things evidence-first — Proven, Probable, Parked, or Pending, no exceptions. Family historian, Airtable nerd, and embracing the AI Squad. I'll be developing further videos on my Youtube channel based on this approach.

The AI Squad

• Evidence Locker: NotebookLM — stores sources, grounds queries, separates fact from assumption

• Map & Timeline Analyst: Gemini — transcribes difficult handwriting from documents accessed through subscriptions, summarizes transcripts, maps routes, and analyses chronology
  Source & Record Expert: Perplexity — record context, verification, and discovery

• Narrative & Ethics Editor: Claude — structures findings, flags when I’m drifting toward assumption or overreach

• Illustrator: Gemini — cover image from verified facts only

Traditional sources in digital form will be consulted along the way — census, maps, parish records, passenger lists, and more.

If you want to go deeper on NotebookLM for genealogy research right now, this post covers my FAN Club setup including a copy/paste prompt: My NotebookLM Setup for FAN Club Research. There will be more on prompts and workflow coming throughout this series.

The Evidence Locker: NotebookLM

The next task is to gather the evidence already known and place it into a structured case file. I have set up a NotebookLM notebook as this evidence locker.

Over the next five weeks, I will be using my AI squad to determine the most promising branches to investigate next and possible records to pursue. I, as case lead, will go and find those records — possibly with the help of Perplexity — and add them (or transcripts) back into the evidence locker for further analysis.

Evidence Type, Purpose in the Case and Status

• Airtable Parish Register Surname Sweep (CSV export) — Extracted parish register entries for surnames amongst Anne Lamb’s descendants (surname sweep), including baptisms, marriages, and burials — In progress

• GEDCOM export — Establish known individuals, relationships, and gaps across four generations of descent from Anne Lamb — Uploaded

• Birth or baptism records — Establish dates, locations and parentage — To be reviewed

• Marriage records — Confirm dates, locations, witnesses, residence, and status — To be reviewed

• Census records — Track household, movement, children, occupation, and social setting — To be reviewed

• Poll books / tax lists — Establish residence, land or property status, and approximate timing for pre-census ancestors, where available — To be reviewed

• Passenger records — Track emigration dates, destinations, travelling companions, and ship — To be reviewed

• Death records — Confirm dates, locations, cause, informant, and official wording — To be reviewed

• Burial records — Identify burial place, religious context, and possible family proximity — To be reviewed

• Newspaper articles — Details about locations, families, associates and more — To be reviewed

• Probate or administration records — Establish locations, family relationships, or absence of property — To be reviewed

• Family stories or notes — Treat as leads, not proof — To be reviewed

Each evidence item can be interrogated individually or alongside the others. Not all items will be fully complete or factual. Some may contradict one another.

Notes for Anyone Following Along

1. The free NotebookLM plan allows up to 50 sources per notebook — plan your uploads accordingly.

2. GEDCOM uploads: I’m going to save you a lot of time here. NotebookLM doesn’t support .ged files directly, and exporting a descendant report can truncate data. The workaround: rename the .ged file extension to .txt — NotebookLM accepts it and the data stays intact. Generating a PDF from a GEDCOM is not recommended; it’s slow and unstable.

3. If you want to see how the Airtable Parish Register Surname Sweep fits into a wider research planning system, this post covers my full Airtable setup for archive visits: My Airtable Setup for Research Planning — Archives and Research Questions

4. Also, Airtable unfortunately doesn’t sync directly with NotebookLM on the free plan yet — a CSV export that you upload as a source works just as well though.

5. Poll books and tax lists: not finding someone in these records doesn’t mean they didn’t live there — many in this family wouldn’t have owned or rented enough property to appear at all. This reinforces Rule 1: mark it unknown rather than guess.

The Rules of the Investigation

• Rule 1: No guessing — If the evidence doesn’t confirm it, it gets a status: Proven, Probable, Parked — or Pending while I wait for more records.

• Rule 2: No single-record conclusions — the case rests on a pattern of evidence

• Rule 3: No sensational language — detective structure is a storytelling device, not a licence for drama

• Rule 4: Historical context supports interpretation — it never substitutes for a source

• Rule 5: The people come before the mystery — not focusing on sensationalism or scandal

Opening Questions

These are the questions I will carry into the next stage of the investigation:

1. Which of Anne Lamb’s grandchildren left Yorkshire — and when?

2. Did they migrate in groups, and did they end up in the same places?

3. Who had close relationships before the migrations? Who appears in each other’s wills and as marriage witnesses — and can I confirm their identities, especially the women?

4. Which parts of this story can be proved — and which must remain unknown?

Case Note: Why AI Belongs In This Investigation

AI will not solve this case for me.

It cannot find a burial entry for me. It cannot decide whether two Henry Dawsons are the same person. It probably cannot tell me why one branch of this family ended up in Illinois and another in Tasmania — and it certainly can’t tell me whether there are more fascinating migration stories hiding in the descendants I haven’t fully researched yet.

But it can help me think.

It can scan through a GEDCOM export covering 702 people and surface patterns I’d miss scrolling through a spreadsheet. I can add additional relevant sources and it’ll scan through those too without blinking. It can flag where my evidence thins out, where we might be assuming rather than proving, and can surface a question I haven’t thought to ask. It can help me organise a timeline, generate search strategies, compare conflicting records, write with more care, and much more.

In this investigation, the AI squad has a clear division of labour. NotebookLM holds the evidence and keeps me moving forward. Gemini helps me read difficult handwriting from documents I’ve accessed through my subscriptions, and maps the geography and chronology. Perplexity finds the record types I don’t know exist yet. Claude structures the findings and supports the other tools. I quite like using the same prompt in two places (Perplexity and Gemini, for example) and asking Claude to sense-check and synthesise the findings.

I remain the genealogist in charge. Every conclusion goes through me. Every Proven/Probable/Parked decision is mine.

That’s the point of the squad structure — not to hand the research over, but to make sure I’m getting the most out of the evidence.

Where the Case Goes Next

In Post 2, I’ll start exposing the gaps systematically — working through Anne’s children and grandchildren branch by branch, establishing what’s known, what’s missing, and where the early records can fill in the picture before civil registration begins in 1837.

Crucially: a family this size, across two marriages, counting four generations of descendants, becomes genuinely fiddly to navigate inside family tree software. Endless clicking between individuals, losing your place, missing connections that aren’t obvious until you suddenly realise at 2am. After 25 years working with computers, I have the repetitive strain injury to prove it.

Querying a GEDCOM export inside NotebookLM is a better way to work with data at this scale. Instead of clicking through 702 people one at a time, I can ask questions and surface patterns. Bring it on.

I'll be sharing the actual prompts I'm using with each tool in the coming posts — what works, what doesn't, and what I'd adjust.

The case file is open. The gaps are already showing. Next Thursday, we start filling them in.

—

Carole McCulloch of NextGenGenealogy is exploring her own brick wall case across the same five weeks. You can follow along at https://nextgengenealogy.substack.com/s/ai-workflow

I had 40+ sources about my brick wall ancestor scattered across folders — county histories, land deed transcripts, migration diaries, researcher emails, church records. So many different (unfamiliar) surnames kept appearing in different contexts, and I knew the answer was in there somewhere. But I was looking at one file at a time, and I couldn’t see the patterns easily. Adding them to my family tree software didn’t feel like a good answer.

Every time I had a new question — “Where did this person live before arriving in Illinois?” or “Who else shows up in these land transactions?” — I’d spend an hour hunting through PDFs, opening files, doing Ctrl+F, trying to remember which document mentioned what. It was exhausting, and I was missing connections.

I needed a way to ask questions across ALL my sources at once.

The Solution: NotebookLM as a Searchable Research Library

NotebookLM is a free AI tool from Google that lets you upload documents and ask questions across all of them simultaneously. It doesn’t search the web — it only searches what you give it. That’s the power: you control the knowledge base.

I uploaded everything I had about my ancestor’s FAN Club — the broad context sources, not individual records — and suddenly I could ask: “Can you find references to the Crosby family?” or “Where does the Bennett family members appear?” and get answers pulled from multiple sources with citations showing me exactly where each reference came from.

I plan to repeat this for each brickwall ancestor. This approach tells me that I have many brick walls I should be able to knock down in months, not years (or at least years, not decades).

My NotebookLM Prompt (Copy This)

Before you start searching through the sources you uploaded, you need to teach NotebookLM how to respond. Think of it like giving instructions to a research assistant.

Here’s the exact prompt I use:

📋 COPY/PASTE PROMPT:

You are a genealogy research assistant assisting another researcher. When I ask questions, provide specific citations to the sources where you found the information. Use a conversational style. If you see conflicting information across sources, tell me what is conflicting and why. Always distinguish between what the sources say and what you’re inferring.

Why this works:

• “Provide specific citations to sources” — Instead of vague answers like “Several sources mention Ohio,” it links you exactly which sources and where

• “Use a conversational style” — Keeps responses readable, not academic

• “ Tell me what is conflicting and why” — Alerts you when sources disagree (critical for genealogy)

• “Distinguish between what sources say and what you’re inferring” — Prevents AI from passing of assumptions or filling gaps as “facts”

You only set this up once per notebook. After that, it remembers.

What to Upload (and What NOT to Upload)

Upload Broad Sources:

• County histories

• Migration narratives and diaries

• Land deed collections or indexes

• Church record indexes

• Biography collections

• Census pages showing your ancestor’s neighbours

• Researcher notes or emails with analysis

• Your Airtable FAN Club synced or exported to Google Drive (if you track individual FAN Club people in Airtable…and I highly recommend it)

• YouTube videos with historical context about a location

❌ Don’t Upload These:

• Newspaper clippings specific to one or a few individuals (track those in Airtable instead)

• Single census records (too granular)

• DNA match lists with full names (focus on DNA match tree names and if you must include the DNA match, change to first name + last initial so they remain private)

• Copyrighted content

• Anything behind a paywall (NotebookLM can’t access it anyway)

The distinction: NotebookLM is for sources that cover multiple people and provide context. Airtable (or a spreadsheet) is for tracking individual records about specific people as structured content.

How to Query Your Sources

Once your sources are uploaded and you’ve set your prompt, you can start asking questions:

Examples of queries that work well:

• “Can you find references to [surname] family?”

• “Which sources mention both [Surname A] and [Surname B]?"

• “Who else appears in connection with [full name]?”

• “What sources mention [place name]?”

NotebookLM will pull results from across all your uploaded sources and show you exactly which document each reference came from. You can click “View Source” to open the full document.

The real power: Cross-source pattern recognition. You’ll see connections between people and places that you missed when looking at files individually.

Notes vs. Sources: Building Your Proof Statement

NotebookLM has four sections:

• Sources (left panel) = Your evidence (the files you uploaded)

• Chat (middle) = Your questions and it’s responses

• Studio = (upper right panel) Pre-built ways to create visual or audio content based on your uploaded content

• Notes (lower right panel) = Your synthesis workspace

When you get back an interesting answer from a query, you can save it as a Note. This is where you start building your proof statement — documenting the connections you’re finding across sources.

Once you’re happy with a note, you can convert it back to a source, making it searchable. So if you later query “Who was connected to Henry County, Ohio?” your synthesis will show up alongside the original sources.

This is hypothesis-driven research in action: write out the connections you think exist, then query your sources to test them.

How NotebookLM + Airtable Work Together

If you’re wondering how this fits with Airtable (or a spreadsheet):

• Airtable = Structured tracking of individuals and specific records (WHO to investigate, WHAT records you have, Proven/Probable/Parked status)

• NotebookLM = Searchable repository of broad context sources (WHAT your sources say about those people)

They solve different problems. Airtable stores detailed information you think is relevant about individuals. NotebookLM stores broader sources and what the historical context says about them — all at once, in one search.

You don’t need both, but if you have both, they’re unstoppable together.

Bonus: If you use Airtable, you can export a table and upload it as a source in NotebookLM. Then you can search across your tracking data AND your context documents simultaneously.

Try This for Your Own Brick Wall

Here’s how to start:

1. Log into NotebookLM

2. Create a notebook specific to your brick wall ancestor

3. Upload sources (county histories, land indexes, migration records — not individual clippings)

4. Set your prompt (copy mine above)

5. Ask your first question: “Can you find references to [surname that keeps appearing]?”

6. Generate timelines

7. Look for patterns — Who appears in multiple contexts? What places connect them?

This isn't a magic genealogy solver. The goal isn’t to find THE answer immediately. It’s to see patterns you’ve been missing when looking at one file at a time. It won't prove relationships or do analysis for you. But it WILL surface the clues you've been missing because they were scattered across too many files.

What’s Next for My Research

I’m still working on the Orin Briggs brick wall case. It’s only been 15 years… <eye roll>

NotebookLM helped me confirm a migration pattern connecting several FAN Club members to a specific county in Ohio before they all appeared together in Illinois. My aunt recently gave me access to her DNA matches, so I’m now testing those connections now too.

I’m not there yet, but I’m closer than I’ve ever been.

I’ll update when I have proof!

Watch the Full Walkthrough

📺 I just published a video showing exactly how I use NotebookLM for FAN Club research — how I set it up, what I found, and how it led to a very exciting (to me) Ohio connection:

📬 Subscribe for case updates as I work through this brick wall (and more NotebookLM workflows as I refine them).

The ancestor I’ve spent the most time on (I’m way too fixated) is this man:

Oren Briggs (born circa 1797 in Massachusetts or Maine, USA and died 1882 and is buried in Cherokee County, Iowa), and he appears in Belvidere, Boone County, Illinois in the 1840 census, working in “manufactures and trades”. (He was listed as a shoemaker in all subsequent censuses but one). The town was founded in 1835 so by 1840, the town was still small enough that everyone knew everyone. By 1853, when Oren and his family left, Belvidere was big enough for mills, schools, multiple churches, lawsuits, land purchases, and overlapping families arriving from Pennsylvania, New York, New England states, and a handful of Midwestern states.

Research into early newspapers, land transactions and other records show a few other people named Briggs, but none indicate any sort of relationship.

Once Oren migrated to Stevens Point, Portage County, Wisconsin, he opened a general store, bought a lot more land, built a large house that is still standing, and even has a street named after him! But by 1858, he’d moved again.

Why I’m pausing research on my strongest candidate in Maine

For a decade, my best lead has been an Oren Briggs of Bradford, Penobscot County, Maine. I’ve done intensive research on this man and his family. He was likely born in Franklin County, Massachusetts, and his parents moved to Penobscot County, Maine not long after the 1800 census. Oren active in his town’s public life through the 1820s and early 1830s. He served as a postmaster, a trader, a town treasurer, and even licensed to sell liquor. He helped Betsey Davis administer her first husband’s estate then married her in 1821, and they had three children together. Then, abruptly, his records shift: lawsuits, friction with local officials, and unpaid debts. By 1834–1835, he disappears from those Maine records altogether, just a few years before someone of the same name and profile begins surfacing in Illinois. His mother, stepfather, brother, and sister-in-law all moved to Henry and Lucas Counties in Ohio around 1835, which is well documented and raises interesting possibilities.

But this is where research discipline matters. As much as I’ve enjoyed exploring the Maine, Massachusetts and Ohio records—and as easy as it would be to keep following that trail indefinitely—I’m setting it aside for now.

To make real progress, I need to stop focusing on one possibility and start understanding the broader community of early Belvidere, Illinois where my ancestor absolutely lived.

Reconstructing the early Belvidere community

To figure out this one man’s origins, I’ve (excitedly) launched a wider project: a full FAN (Friends, Associates, Neighbours) reconstruction of early Belvidere from the town’s establishment in 1835 until Oren’s departure in 1853.

This isn’t quite a formal One-Place Study, but it edges close. I’m creating a structured dataset of every identifiable settler in this period—where they were born, where they came from, who they married, all so that I might figure out these migration routes that fed into Boone County so that I learn where Oren stopped off prior to Belvidere and possibly who he could have traveled with.

If those records were online, I’d also include full land-transaction networks, but many of them require on-site research.

The aim is simple: When an ancestor’s origins are unclear or incomplete from records, the community around them often provides the missing context.

Oren almost certainly didn’t arrive in Illinois directly from Massachusetts or Maine. I keep seeing places in New York State pop up. These origins form patterns, and it’s those patterns that help anchor someone whose own paper trail is incomplete.

Why this broader approach matters

I’m taking this route because Oren’s Illinois records don’t offer the answers I need, but the wider settlement might.

The Maine man remains the strongest possibility, yet I can’t connect the two with confidence until I place Oren inside a clear migration structure—regional ties, shared origins, neighbour networks. I want an answer that can be proven.

And because I naturally focus on migration patterns, this approach fits. It lets me look at movement, context, and how early communities formed and shifted across the region.

A first look at the migration map (v1)

I’ve generated a Version 1 migration map, using NotebookLM, which I’ll refine as the dataset grows. Right now, I know it’s rough around the edges. Even at this early stage, patterns are emerging: families who travelled through predictable New England channels, others who moved west in larger leaps, and a few whose origins overlap intriguingly with Oren’s possible Massachusetts–Maine background.

It’s early, but promising.

Shoutout to who explains her process of generating NotebookLM infographics at https://substack.com/home/post/p-180919923.

Core steps in my plan

• Finish building a master name list for Belvidere and immediate surroundings using the 1840 and 1850 federal censuses (already done), and 1855 state census, cross referencing names pulled from existing indexes and compiled sources first: census indexes, Boone County biographical and history volumes, cemetery indexes, and the major Boone County index compilations (including probate, land, and vital indexes).

• Assign a priority ranking based on “proximity” to Oren or otherwise try to group people into FAN clusters—possible family members, close neighbors, business partners, land transactions, co‑witnesses on deeds, fellow church members—and track those networks over time to see how Belvidere’s community formed and evolved.​ I definitely see the name names coming up time and time again.

• For the highest ranking people/households, follow them across record sets: county land and deed books, tax lists, probate and court records, church registers, cemetery transcriptions, newspaper items, and local directories where they exist.​

• I’ll note individual origins (where possible) and places they passed through (birth locations of children)

• When possible, note where early Belvidere families moved on to Wisconsin (the border is very close), further west or back to their home state

• Keep the Massachusetts, Maine & Ohio material in reserve while letting the community context develop

How this differs from existing Boone County work

• Standard Boone County histories (for example, the ILGenWeb history page, Genealogy Trails’ Boone County narratives, and 1912 The Past and Present of Boone County, Illinois) highlight selected pioneers and prominent citizens but do not attempt a complete reconstruction of every resident.​

• My project is deliberately exhaustive and record‑driven: the goal is not just to tell stories about a few leading figures but to document the town’s early population and its web of relationships in a way that can eventually support descendants, one‑place researchers, and migration studies.​

Key online sources I plan to build on

• Boone County, Illinois Genealogy (FamilySearch Wiki) – overview of records and repositories.​

• Boone County, Illinois – ILGenWeb history and records pages.​

• Boone County, Illinois History & Genealogy (Genealogy Trails), including pioneer‑life texts and transcribed histories that name early settlers.​

• The Past and Present of Boone County, Illinois (digitised on Internet Archive) for township histories and biographical sketches.​

• Digitised historical encyclopedia and similar Boone County volumes with chapters on early surveys, land titles, and township history.​

• I can also contact the Winnebago & Boone Counties Genealogical Society; they have done research for me in the past.

What I expect this project to reveal

There’s a good chance that Oren’s origins will become clearer once I understand the people he lived among.

And even if the answer is more complicated—or remains elusive—I’m building a richer picture of early Belvidere’s formation, which is still interesting since I have various ancestors that passed through there.

Follow along!

• If you have worked on similar migration clusters, I’d love to hear from you.

• Follow this Substack for serialised case studies of specific Belvidere FAN clusters and methodological walk-throughs.​

• Or if you have ancestors in Boone County, Illinois during 1835–1853. Shared patterns often reveal more than any single record can. Contribute scans, family stories, or pointers to obscure record sets related to pre‑1860 Belvidere families, which can be integrated into the town‑wide FAN reconstruction.​

If you’re curious where my interest in migration starts, I’ve moved from the US to the UK twice, and I’ve also written about the past 18 months of full-time travel shaped by burnout in perimenopause and a late-in-life neurodivergent diagnosis. That reflection is on my other Substack, if you’d like to read it.

I’m focusing on handwritten records, because that’s where the real gains are. I can read English and American handwritten documents back to the 1700s and sometimes even earlier, but it takes a lot of studying and practice. More on that in a bit.

So I’m running documents where I get stuck (or there are many of them) through TryLeo AI, and it has transformed the way I work. (By the way, there’s a free tier. If you want to use my TryLeo affiliate link, we’ll both receive 50 lifetime credits).

What I’m Transcribing

• Deed indexes and deeds.

• Court ledgers.

• Wills.

• Census records.

• Church records.

• Anything else that’s been handwritten or typed :)

Fully-typed documents are easy enough for NotebookLM or ChatGPT to handle on its own, but any scribbled, handwritten or low-quality image goes straight into TryLeo.

Where TryLeo Excels

Its accuracy on messy handwriting has surprised me, and I’ve been able to finally read sections I had previously skipped because I simply couldn’t make out the words. It’s great at identifying sections that have been written vertically as well as horizontally.

It’s good for printed forms and printed or typed documents. Great for English documents when the handwriting is GOOD and good, clear, high-resolution images. I like the organisation into folders. So far I’m copying the same folder structure as my genealogy research files.

I love that you can make your edits to the auto transcription, save them as draft, export as text or PDF, etc.

I think of it as a quicker way for me to get to a transcription. It’s doing the bulk of the work but you still get the excitement of going through line by line.

Where TryLeo Falters

I am having mixed results on my German handwritten documents.

TryLeo transcribed the above as:

Wilfalen Jöhnen Arthur Osten, Dr. med. und grach. Arzt und Iffner Chf Frau Voglin P. Wilfalenius Lucilia geb. Schroeder, Hannover, Tintarichtstr. 1.

ChatGPT was surprisingly better:

Wilhelm Johannes August
Osten, Dr. med. und prakt.
Arzt und Frau Sophia
Maria Wilhelmine Emilie
geb. Schroeder, Hannover,
Pintorffstr. 1.

I am fully aware there are potentially still mistakes being introduced and so I don’t consider AI a total replacement for knowing how to read the document directly, plus the historical context of the record, etc.

TryLeo also skipped over dates in a census record, and ChatGPT had a look at the document and guessed. They were WAY off.

Playing around with NotebookLM

I’ve been using ChatGPT to organise transcripts, and recently NotebookLM too, because it ‘hallucinates’ much less than ChatGPT (which is still brilliant in other ways).

From there, I’m using the transcripts to generate:

• Abstracts.

• Timelines.

• Structured research plans.

• Clean CSV files ready for Airtable.

Airtable is where I track research clusters, genealogy questions, migration notes on each person by source, compare names across counties, and see connections that don’t always surface in traditional family tree software. It’s also ideal for name-matching when I’m working on people who appear repeatedly across deeds or court documents.

Where a Human Genealogist Is Still Needed

A good transcription isn’t the same as a good conclusion. AI can misread unfamiliar surnames, flatten archaic spellings, or mistranslate idioms. It can’t judge whether you’re looking at the right John Smith. And when you’re working across languages, translation tools get you 80 percent of the way, but the final 20 percent is often where the genealogical gold is found.

And as we’ve seen, it not totally accurate (yet). I’m still learning how to read old German handwriting (I definitely recommend Germanology Unlocked courses) so that I can be more confident of the record contents.

Just like the online indexes, AI can still spit out errors. I treat AI output as a draft. The interpretation, the linking, the pattern-spotting? That’s still the work of a researcher.

Why This Matters for a Genealogy Do-Over

If you’re doing an audit or starting fresh, you need every detail. But many of us skim the hard documents because the handwriting feels impossible. That means we miss clues about relationships, land boundaries, witnesses, neighbours, and debts. Transcription is the foundation of accurate work. And with the right tools, it becomes fast enough to keep up with your research.

My next project is using AI to compare a lot of deeds. It should be able to match or a least identify neighbouring boundary descriptions.

Technology can’t choose the right documents for you. That part still relies on human judgement. But once you have them, this workflow removes hours of manual work and lets you focus on analysis.

How are you using AI?

I’m building prompt libraries and multi-step workflows that turn these transcripts into organised research tools. But I realise now they are prone to change because the technology is moving so fast. Before I work on these more, I’d love to know:

• Which documents do you struggle to read?

• What’s the part of the process you get snagged on?

• Would you like a step-by-step walkthrough of my TryLeo → ChatGPT → Airtable workflow?

Thanks all!

P.S. If you want to try TryLeo yourself, here’s my affiliate link. And if you have questions, please leave them in the comments — they’ll help me shape the next post.