Timelines are essential in family history research, but creating them manually is time-consuming. You need to gather data, enter it into tables, and organise everything chronologically.
AI can help with this process. It extracts data from your family history documents, analyses it, creates tables, and orders events by date.
I’ve tested this approach to discover what works and what doesn’t. This post shares my findings and shows you how to get started with simple genealogy timelines. Later posts will examine more complex timelines for research problem solving and discuss in more detail how to optimise the prompt and input data for best results.
Why Use Genealogy Timelines
Timelines place people in time and space. They help you:
- Clarify identities and solve research problems
- Expose gaps and inconsistencies in your data
- Organise information for better research planning
- Create engaging family narratives.
If you’re not using timelines yet, read my article Use Genealogy Timelines to Organise, Analyse and Improve Your Research for a discussion of the benefits and uses for genealogy timelines.
The Timeline Problem
Online family trees and family history software create limited timelines. You might get a timeline for one person with their immediate family, but events for the family are limited to birth and death. The format is usually PDF-only, which means no editing in Excel.
This limits their value for family history research. You have limited control over what data gets included and can’t make adjustments afterward.
Creating custom timelines manually means extracting data from your software, entering it into spreadsheets, and organising it chronologically. It’s tedious work that’s prone to errors.
What AI Can Do for Timeline Creation
AI excels at three key tasks:
- Extract data from documents and format it into tables
- Organise events chronologically to create proper timelines
- Convert results into Excel-compatible formats.
AI can also provide historical context and suggest research directions, but I’ll focus on the core timeline creation process here.
Before You Start: Essential Preparation
Protect Privacy First
Always exclude living people from any data you share with AI tools. Check your software settings and review reports to ensure no personal information about yourself or living relatives gets included.
Decide Your Timeline Content and Format
Consider what information you need:
- Dates: Full dates or years only? How should approximate dates (abt. 1832) or spans (1841-1844) be handled?
- Events: Birth, death, marriage are standard, but baptism and burial are useful too. Include residence, or occupation events?
- Names: Full names in one column or separate surname column?
- Locations: Complete location in one column or split into separate columns?
- Citations: Useful, but I left them out in these simple timelines.
I also recommend including a unique identifier column to distinguish between people with the same name.
For more tips on genealogy timelines, see my article Excel Genealogy Timelines: Complete How To Guide.
Testing Different AI Tools and Formats
I tested three AI tools with different document formats:
AI Tools Tested:
- ChatGPT 4o (paid)
- Google Gemini 2.0 (free)
- Claude Sonnet (paid)
Document Formats Tested:
- Ancestry individual profiles (PDF and text)
- Legacy Family Tree individual reports (PDF)
- Legacy Family Tree family group sheets (PDF)
The Prompt That Works
Here’s the prompt I developed through testing:
‘Please extract all birth, baptism, marriage, death, burial, and residence events from the following report. Include only events where dates and locations are explicitly stated – no assumptions. Present results in a table with four columns: Date, Type of Event, Location, and Name of the Person. Sort chronologically by date. Omit any index or row numbers. If a date is approximate (e.g. ‘Abt.’, ‘Bef.’, ‘After’), retain the original text but sort chronologically based on interpreted value.’
For software-generated reports with unique identifiers, I added:
‘Please add another column called RIN. This is the unique identifier number for a person. You will find that number in square brackets after a person’s name. Add the RIN for each person in this new column.’
Results: What Worked and What Didn’t
| Document Type | ChatGPT 4o | Google Gemini | Claude Sonnet |
| Ancestry Individual profile PDF | Failed | 100% Success | 100% Success |
| Ancestry Individual profile Text | 100% Success | Not tested | Not tested |
| Legacy Individual profile PDF | 100% Success | 100% Success | 100% Success |
| Legacy Family Sheet PDF | 100% Success | 100% Success | 100% Success |
Key Findings:
Format matters. Image-based PDF documents and PDF documents from web pages (like Ancestry) can be problematic. They can be difficult to search; have overlapping elements or graphics; and may include OCR (Optical Character Recognition) text that is less accurate. ChatGPT struggled with these and could only interpret the data when it was converted to text format by copying and pasting.
John Beaumont has a great video about PDF issues and AI, if you need to know more.
Structure helps. Reports from family history software work best because they’re already consistently formatted.
Instructions are crucial. AI tools interpreted standard events (birth, death, marriage) accurately but needed guidance with residence events that appeared in different formats.
Example additional guidance provided:
‘You have not listed any residence events. I think you had trouble because they have been expressed in a few different ways. There are events in the document that say ‘resided at [address] in [year] in [location]’; others say ‘had a residence in [year] in [location]’. Can you see if you can extract those ones and add them to the timeline?’
Google Gemini performed particularly well, correctly interpreting events as residence events even where the word residence was not mentioned. For example, it assumed that a religion event with a date and location indicated residence.
Quality Control: Review Your Results
Always check AI-generated timelines, even when they look perfect. Common issues include:
- Minor errors: Events slightly out of chronological order
- Major errors: Missing events, wrong people, or invented information.
ChatGPT’s failure with the Ancestry PDF was spectacular. It invented people, changed locations, and created fictional events and people. When I pointed out errors, it created even more problems. This shows why testing different AI tools matters.
Quick Tips for Success
- Start simple with individual or family group reports
- Use structured data from family history software when possible
- Test different AI tools if one doesn’t work well
- Review results carefully before using the timeline
- Refine your prompt based on your specific data format.
What’s Next
Simple timelines work well with AI, but what about complex family research? In Complex Genealogy Timelines with AI, I explore multi-generational timelines and solving the errors that arise with more complex data.
The key takeaway: AI can dramatically speed up creation of genealogy timelines, but success depends on preparation, the right tools, and careful review of results.
About the Author
Danielle Lautrec is a genealogy educator, researcher, and author of The Good Genealogist. With qualifications in history, family history, and historical archaeology, she teaches for the Society of Australian Genealogists.