FamilySearch Labs Full-Text Search:
The Tool That Changes Everything
For the first time in history, you can search every word of a handwritten land deed or probate record — not just the indexed names. Here is what it is, how it works, and how to use it today.
There is a frustration known to every genealogist who has ever worked with land records or probate files: the index only tells you part of the story. The deed index names the grantor and the grantee. The probate index names the deceased and the executor. But inside those handwritten documents — tucked into boundary descriptions, witness lists, and margin notes — are dozens of other names: the ancestor's neighbours who witnessed the deed, the siblings who contested the will, the in-laws who appeared as sureties, the children listed as heirs. None of those names were indexed. None of them were searchable. Until now.
FamilySearch Labs Full-Text Search is one of the most significant developments in genealogy technology in decades. Using artificial intelligence trained to read historical handwriting, it has made the complete text of nearly two billion document images searchable by keyword — for free, with no subscription, no login required beyond a free FamilySearch account. If your ancestor's name appears anywhere on any page — as a witness, as a neighbour, as a margin note, as a boundary landmark — this tool will find it. This guide will tell you exactly what it is, how it works, and how to start using it today.
Traditional indexing captures perhaps a dozen names per document. Full-Text Search makes every word searchable. For land and probate records, that is the difference between finding one person and finding an entire neighbourhood.
🔬 What FamilySearch Labs Is — and What "Experimental" Really Means
FamilySearch Labs is the innovation testing area of FamilySearch.org — the place where new features are made available to regular users before they are formally released across the whole platform. Think of it as the front porch of the main house: fully accessible, genuinely useful, but still being refined before it moves inside permanently.
The word "experimental" can make nervous users hesitate — it sounds unstable or incomplete. In the case of Full-Text Search, it simply means the tool is still being improved based on user feedback. The core functionality is robust and has been used by hundreds of thousands of researchers since its announcement at RootsTech 2024. The genealogical community labelled this technology a "game changer" immediately upon its announcement, and for good reason — the tool aims to make over a billion previously difficult-to-access records fully searchable, records that were either unindexed or only partially indexed.
Being an experimental feature means two practical things: first, the interface may change as FamilySearch improves it, so occasional minor differences from the screenshots described in this guide are normal. Second, FamilySearch actively wants your feedback — there is a built-in feedback mechanism in the tool, and researchers who report errors or suggest improvements are genuinely contributing to making it better for everyone.
📋 The Problem It Solves: What Traditional Indexing Misses
To understand why Full-Text Search is revolutionary, you need to understand the limitation it overcomes. Every genealogy database — FamilySearch, Ancestry, MyHeritage, and all others — traditionally presents records through an index: a list of searchable fields that human volunteers or automated systems have extracted from each document. For a land deed, the index might contain the grantor's name, the grantee's name, the county, and the year. For a probate record, it might contain the deceased's name, the executor's name, and the filing date.
- Searches only the extracted fields — typically 2–4 names per document
- Misses witnesses, neighbours, and heirs listed within the document text
- Cannot find an ancestor mentioned as a boundary neighbour in a land description
- Cannot search for relationship terms like "my son-in-law" or "widow of"
- Cannot find place names mentioned within the document body
- Millions of records remain entirely un-indexed and invisible to search
- Searches every word on every page of every covered document
- Finds witnesses, heirs, bondsmen, sureties, and neighbours by name
- Locates ancestors mentioned in boundary descriptions ("bounded north by John Murphy's land")
- Searches for relationship phrases like "my daughter" or "late husband"
- Finds township names, waterway names, and plantation names
- Makes unindexed collections fully searchable for the first time
For example, a probate file's index might list only the deceased and executor, or a deed index only names the grantor and grantee — leaving out witnesses, heirs, neighbors, and other names mentioned in the documents. FamilySearch's Full-Text Search tackles this problem using AI-based handwriting recognition and text recognition technology. In simple terms, the system scans every page of a document image and "reads" all the words, whether typed or handwritten, and makes them keyword-searchable.
🤖 How the AI Actually Works: HTR Explained Simply
The technology powering Full-Text Search is called Handwritten Text Recognition, or HTR. It is related to — but significantly more advanced than — the optical character recognition (OCR) that has been used to make printed text searchable for decades. The difference is important.
🚀 Quick Start Guide: Accessing Full-Text Search in Six Steps
The following guide works as of May 2026. FamilySearch occasionally updates the Labs interface, but the core pathway — Labs page, find the experiment, enable it, search — remains consistent. If the layout has changed slightly since publication, look for the "Full Text" or "Expand Your Search" option in whatever Labs menu is currently visible.
Year range: 1850 – 1900
Place: County Cork or Illinois
Collection: (optional — narrow to land records or probate)
🗂️ What Records Are Covered in 2026
The scope of Full-Text Search has expanded dramatically since its launch. The initial focus was on US Land and Probate Records and Mexico Notary Records, but over a billion record images have been added since RootsTech 2024. Major collections now include US Legal, Vital Records, Migration Records, Land/Probate, and Military records. As of early 2026, the collection stands at nearly two billion searchable images.
As of January 2026, nearly 2 billion record images are searchable, spanning U.S. land records, Mexican notarial records, Brazilian civil registration, Canadian documents, UK records, and hundreds of previously browse-only collections.
| Record Type | What's searchable within the text | Available in 2026? |
|---|---|---|
| US Land & Deed Records | Grantor, grantee, witnesses, boundary neighbours ("bounded north by John Smith's land"), legal descriptions | ✔ Yes — core collection |
| US Probate Records | Deceased, heirs, executor, creditors, inventory items, witnesses to will, relationship terms ("my daughter," "late husband") | ✔ Yes — core collection |
| Mexico Notarial Records | All parties named in notarial acts, property descriptions, witnesses | ✔ Yes — core collection |
| US Military Records | Service member, dependents, witnesses, pension correspondence | ✔ Yes — added 2024–2025 |
| UK Records | Varies by collection; wills, land, and civil registration documents | ✔ Yes — growing collection |
| German, French, Italian, Dutch | Church, civil, and land records in these languages | ✔ Expanding in 2026 per FamilySearch announcement |
| US Federal Census | Every name on every page | Not yet — indexed separately; watch Labs for updates |
FamilySearch will expand its full-text search technology in 2026 to read old handwriting in additional languages, making newly digitised documents easily searchable by an ancestor's name. Records are currently available in English, Spanish, and Portuguese, with plans to add Chinese and other European languages such as French, German, Dutch, and Italian.
🧠 Seven Search Strategies That Get Results
Full-Text Search rewards creativity. Because it searches the entire document — not just named fields — you can look for your ancestor through many different angles simultaneously. Here are the strategies that experienced researchers use most effectively.
Search the surname alone first, then narrow
Start with just the surname in the keyword box and apply a date range and state filter. Review the first page of results — if there are too many, add the given name. If there are too few, remove the date filter. Build from broad to narrow.
Use wildcards for spelling variations
The asterisk * replaces multiple characters. Mull* will find Mulligan, Mullen, Mullin, and Mullery. Use this for surnames with common spelling variants or frequent transcription errors. The question mark ? replaces a single character.
Search for relationship phrases, not just names
Try phrases like "widow of", "my daughter", "late husband", or "heirs of" combined with the surname. A will that says "to my daughter Margaret, wife of Patrick Sullivan" appears in a search for Sullivan daughter even if Margaret's maiden name is what you are trying to find.
Search for boundary neighbours and landmarks
One of the most powerful ways to use FamilySearch AI search is by seeking landmarks rather than people. By searching for specific waterways, mountains, or neighbouring plantation names, you can recreate an entire neighbourhood. Try searching [creek name] + [county] to find all deeds describing property along that waterway.
Search for your ancestor as a witness, not just a party
Traditional indexing rarely captures ancillary mentions, yet they can provide crucial clues in genealogical research. With AI Full-Text Search, these previously invisible mentions become findable. An ancestor who witnessed a neighbour's deed is now discoverable — confirming they were in that location in that year.
Filter by specific collection
AI helps group records into collections using the metadata gathered when the records were digitised. Use the collection filter to limit your search to land records, probate records, or military collections. This dramatically reduces irrelevant results when you already know which record type you are looking for.
Search for women through their husbands' records
Women can be located more frequently in FamilySearch Full Text records — mentioned as daughters in a father's will, as a witness, or as a wife waiving her dower rights in a land sale. By searching the full text of documents, we can finally find the traces of females who appeared in traditional records.
🔦 A Worked Example: The Brick Wall That Fell in 20 Minutes
The following example, based on the experience of a real researcher, illustrates how Full-Text Search finds ancestors that traditional index searches miss entirely.
The problem: A researcher was searching for the father of John Robert Dyer, born about 1813 in Hawkins County, Tennessee. Traditional searches on Ancestry and FamilySearch turned up very few Dyer men in that county as candidates, partly due to local record loss.
The Full-Text Search: The researcher entered "Dyer" into the Full-Text Search box and filtered to Tennessee and the relevant time period. Within minutes, an Ignatius Dyer appeared — not as the primary party in a deed, but mentioned in a store account book from Rogersville, which sits in Hawkins County.
The result: Ignatius Dyer was listed in a Rogersville store account book, held by the Tennessee State Library and Archives. A broad search on Ancestry and FamilySearch had not previously surfaced any other records of Ignatius Dyer. The store account book — the kind of document that would never appear in a traditional name index — had been transcribed by the AI and was now searchable.
✔ Brick wall broken in one search session⚠️ Honest Limitations: What the AI Cannot Do
Full-Text Search is genuinely revolutionary. It is also imperfect, and understanding its limitations is as important as understanding its capabilities.
The AI struggles with the flourishes of seventeenth-century colonial script, ink bleeds, or damaged paper. Very old documents, water-damaged pages, and unusual regional handwriting styles produce more errors. Names are particularly prone to transcription mistakes — a letter misread in the middle of a surname can make the document unsearchable by that name. This is why wildcard searches are valuable, and why reading the original image alongside the transcript is always essential. The transcript is a tool for finding documents. The image is the evidence.
Three other limitations worth knowing: First, not every FamilySearch collection is covered. If a collection has not yet been processed by the HTR system, it will not appear in Full-Text Search results — it may still be browseable or traditionally indexed, but not full-text searchable. Check the collection filter to understand what is currently covered. Second, AI-generated summaries are often provided in English even for foreign-language documents — which is helpful for English-speaking researchers but means the summary may occasionally miss nuances in the original language. Third, the system is constantly expanding and improving: a collection not covered today may be fully searchable in three months. Returning to Full-Text Search periodically with searches you have previously run is always worthwhile.
✦ Why This Changes Everything for Brick Walls
The genealogical brick wall has always been, at its core, an evidence problem. Your ancestor existed — there is no doubt of that. The question is whether the evidence of their existence survived, and whether you can find it. For a century, the answer to the second part of that question was limited by what human indexers had the time and resources to extract from the documents. Full-Text Search removes that limitation.
Consider all the ways an ancestor might appear in historical records beyond just being the primary subject. Traditional indexing rarely captures these ancillary mentions, yet they can provide crucial clues in genealogical research. With AI Full-Text Search, these previously invisible mentions become findable. Records can be effectively "lost" within archives due to various practical issues. Even when digitised, these records remain essentially hidden if the catalogue metadata is incorrect. AI Full-Text Search, however, scans the actual document content, allowing it to locate records regardless of cataloguing errors.
The ancestor who witnessed a deed in 1847 and was never indexed. The woman who waived her dower rights in her husband's land sale and appeared only in the body of a legal document. The grandfather whose name appears as a boundary neighbour in six different deeds, none of which listed him in their index. All of them are now findable. All of them are waiting in a search box at familysearch.org/search/full-text.
It is free. It covers nearly two billion documents. And it is getting better every month.
Your Next Breakthrough
Is Already Searchable
Open FamilySearch Labs and run a Full-Text Search for the surname you have been stuck on. Search as a witness. Search as a boundary neighbour. Tell us what you find — breakthroughs, surprises, and even the searches that need more work. Every result teaches us something.
Open FamilySearch Labs →