Testing Handwriting OCR for Genealogy: Transcribing Death Records with AI

AI Handwriting OCR for Genealogy: Death Record Transcription Test and Review

As genealogists, we know the value—and the challenge—of working with historical documents written in long-forgotten hands. Whether deciphering ornate script or interpreting fading ink, extracting the data we need from original sources takes time, patience, and a good eye. That’s why I was intrigued when I was invited to test Handwriting OCR, an AI-powered handwriting-to-text converter that promises to transform even the toughest handwriting into editable, searchable text with speed and precision. 

In this blog—the first in a new series where I’ll test this platform across different historical document types—I’ll be sharing my results and observations from using Handwriting OCR on two death certificates from my own family history collection. My goal? To evaluate how this tool performs when handling one of the most commonly used—and often difficult to read—records in genealogical research.

Why Death Certificates?

Death certificates are critical genealogical records. They typically include a wealth of information: full name, date and place of death, age at death, cause of death, and often key clues like parents' names and birthplaces. But they also present unique transcription challenges. The handwriting is often hurried, medical terminology is complex, and formatting varies by time and location.

Most existing OCR platforms struggle to process these handwritten documents effectively, so testing death records seemed like a fitting place to begin.

Case Study 1: Samuel Steinman, 2nd Great-Grandfather

AI Handwriting OCR- Death Certificate Example 1

*Note: Click to zoom in

AI Handwriting OCR- Death Certificate Example 1-2

AI Handwriting OCR- Death Certificate Example 1-3

AI Handwriting OCR- Death Certificate Example 1-4

Handwriting OCR did a surprisingly competent job transcribing this document, but several subtle errors did appear—mostly understandable given the script and formatting challenges. Here are the most noteworthy findings:

• Date of Birth: Transcribed as 1853 instead of the correct 1855. The final digit on the certificate was disconnected, likely making it difficult for the AI to interpret.
• Mother’s Maiden Name: Interpreted as Snell instead of Sherban. This was partially obscured by a descending letter “g” from the line above—something even a human might misread without close inspection.
• Mother’s Birthplace: Listed as Bern, Germany, but should have been Pennsylvania. The AI likely misread “Penn.” as a foreign place name due to unclear abbreviation.
• Cause of Death: “Pyelonephritis (Duplex)” was misread as “Pyelonephritis (Angina)”—again, a nuanced medical term likely outside the AI’s contextual vocabulary.
• Undertaker’s Name: The name was difficult even for me to decipher, and the AI’s guess (Hutson & Son) was close to what I initially read (H— & Son), though not definitively correct.
• Form Numbering and Formatting: Some inconsistencies in how form elements were grouped, but overall structure was retained.

​Final Thoughts on Samuel’s Record:
Despite a few small inaccuracies, this was an impressive result for a handwritten death certificate. These are the kinds of records that most OCR platforms simply can’t handle at all, and here the AI provided a nearly complete transcription with minimal editing needed. A human still needs to verify every detail (as we always should), but this tool significantly reduces the time spent manually transcribing.

Case Study 2: Mary Harriet (Hulet) Steinman, 2nd Great-Grandmother

AI Handwriting OCR- Death Certificate Example 2

*Note: Click to zoom in

AI Handwriting OCR- Death Certificate Example 2-2

AI Handwriting OCR- Death Certificate Example 2-3

AI Handwriting OCR- Death Certificate Example 2-4

Mary Harriet’s certificate introduced different layout and formatting challenges, including more structured government fields and handwritten entries nestled between printed form labels. Here’s how Handwriting OCR handled it:

• Location Details: The City was mistakenly placed in the Township line of the transcript, which could confuse localities in more complex regions.
• Section Headers: The Medical Certificate of Death title was moved from its intended place in the transcript, but all content beneath was still included.
• Demographics Block: The AI jumbled categories like sex, race, and marital status together (“female White Widow”), though it transcribed the handwritten answers accurately.
• Date of Birth: The document showed 1852, but the transcription read 1857—a five-year discrepancy due to misreading a faint final digit.
• Age at Death: The handwritten form listed 87 years, 4 months, and 11 days, but the OCR compressed this line and read the months as 11, likely due to formatting inconsistencies.
• Father’s Name: Interpreted as Jas. Hulitt instead of Ira Hulett. The script was admittedly difficult here, and the misreading was understandable.
• Filing Details: Slight errors in the month and clerk’s name, which were extremely faint and partially disconnected.
• Physician Information: Misread initials, address, and timeline of medical attendance—small but notable for researchers validating cause of death.
• Document Numbering: A few number sequences were clumped together, but data was not lost—just slightly misstructured.

Final Thoughts on Mary Harriet’s Record:
Again, I was impressed. Despite some layout issues and detail-level inaccuracies, Handwriting OCR extracted nearly all the handwritten content with a high degree of readability. For genealogists, these transcripts are immensely helpful starting points, especially when working with challenging handwriting or in bulk.

Takeaways from Testing Death Records on Handwriting OCR

AI-powered transcription tools like Handwriting OCR are opening exciting new doors in genealogical research. While they are not perfect (nor should we expect them to be), what they offer is a huge leap forward in automating some of the more tedious aspects of our work. Here’s what stood out most in this first test:​

• ✅ Strong transcription accuracy, even on cursive and challenging handwriting
• ✅ Minimal formatting loss, with most sections retained and recognizable
• ✅ Easy export features, which I’ll explore in more detail in future tests
• ❗ Minor transcription errors, typically involving faint or disconnected text
• ❗ Contextual misunderstandings, particularly with abbreviated place names or medical terminology
• ❗ Structure inconsistencies, in category grouping or field labeling​

​Yet, in both of my tests—death certificates from the same location in Gentry County, Missouri, but from different decades—Handwriting OCR provided complete, editable versions of complex historical documents that would normally require slow, manual entry.

For any genealogist who’s ever squinted at an old death certificate under poor lighting or spent hours transcribing by hand, this platform offers a welcome solution. It’s not a replacement for human verification, but it’s a powerful companion that can accelerate our workflows and help us focus more on the analysis, stories, and connections that make genealogy so rewarding.

Coming Soon: Testing Handwriting OCR on Other Genealogy Documents

This is just the beginning. In upcoming posts, I’ll be testing Handwriting OCR’s capabilities on:

• Census records
• Marriage certificates
• Old letters and family Bibles
• Naturalization and immigration records
• Church registers and more

Stay tuned as I continue exploring how this platform holds up across the variety of handwritten documents we work with every day. If you've tried Handwriting OCR on your own records, I’d love to hear about your experience!
​
Have you tested AI handwriting tools for genealogy? What worked (or didn’t) for you? Let me know in the comments!

If you find that this would be a valuable addition to you genealogy research toolkit, Handwriting OCR will provide an exclusive offer code to all my blog readers! Just let me know!

​​More AI and Genealogy Resources

Learn about more AI in Genealogy and other family history and genealogy resources under the AI in Genealogy and  Genealogy Resources category and on my dedicated Genealogy Resources page.