Machine Learning Engineer Co-op

Testing Ancestry's new AI "Fill in Timeline Gaps" Feature: my first impressions. Artificial intelligence continues making its way into genealogy, and Ancestry has quietly introduced another AI-powered feature that caught me completely by surprise. About a week and a half ago, while I was in the middle of researching one of my ancestors, I noticed something new. As I worked through an ancestor's profile in the Facts view, I saw small "Description" buttons appearing next to some - but not all - of the timeline facts. Curious, I clicked one to see what it was. That led me to Ancestry's new BETA "Fill in Timeline Gaps" feature. Of course, I had to stop what I was doing and put it through a few tests. What is the "Fill in Timeline Gaps" Feature? When you click one of the new Description buttons, a side panel opens displaying Ancestry's BETA Fill in Timeline Gaps tool. Ancestry AI Fill in Timeline Gaps Feature- Example The feature uses AI to generate a description for a specific timeline fact. Before generating the text, you're given three options for how detailed you'd like it to be: * Just the basics * More detail * In-depth The generated description can then be added directly to that individual fact. It's a pretty simple concept, but one that has the potential to make ancestor profiles much more informative - especially for users who don't normally add descriptions to their timeline facts. Ancestry AI Fill in Timeline Gaps Feature- Detail Choices Example Why fact descriptions matter. Many Ancestry users simply enter a date and place for events like births, residences, marriages, or deaths and move on. While that's perfectly acceptable, the Description field provides an opportunity to add context that helps tell the story behind each event. For example, instead of simply recording a residence, you might note that the family was living with relatives, renting a farm, or appearing in a particular census enumeration district. A birth description might include details taken directly from a birth record, while a military fact might summarize information found in a pension file. Adding these details can make an ancestor's timeline much richer and easier to understand without having to open every attached source. That's exactly why I was interested to see how well Ancestry's new AI handled this task. My first test. The first person I tested was my fourth great-grandfather, Nehemiah Hulet, simply because I happened to be working on his profile when I discovered the feature. I selected the Birth fact and generated all three description options. This test actually impressed me. The AI stayed focused on the birth event itself. * Just the basics provided a short, straightforward description. * More detail expanded on the information while remaining focused on the birth. * In-depth added additional context without wandering into unrelated parts of his life. Each level simply built upon the previous one, exactly as I expected. So far, so good. Ancestry AI Fill in Timeline Gaps Feature- Example 1 | Ancestry AI Fill in Timeline Gaps Feature- Example 1-2 | Ancestry AI Fill in Timeline Gaps Feature- Example 1-3 | My second test produced very different results. Next, I switched to my "official tester ancestor," my second great-grandfather, Samuel Steinman, whose tree I've often used to evaluate new genealogy features. Once again, I selected the Birth fact and generated all three levels. This time, however, the results were completely different. Instead of describing Samuel's birth, the AI began summarizing his life. * Just the basics produced two sentences that briefly summarized his life. * More detail expanded those summaries into two longer sentences. * In-depth generated a four-sentence mini biography. The information wasn't necessarily incorrect, but it didn't belong in the description field for a single birth event. Rather than expanding upon the birth fact itself, the AI had shifted into writing a short life summary. That wasn't what I expected the feature to do. Ancestry AI Fill in Timeline Gaps Feature- Example 2 | Ancestry AI Fill in Timeline Gaps Feature- Example 2-2 | Ancestry AI Fill in Timeline Gaps Feature- Example 2-3 | Testing a residence fact. I decided to keep experimenting. I selected one of Samuel's Residence facts - one of only a couple that displayed the new Description button. This time, the results were much closer to what I had hoped for. The AI focused on the specific residence and even incorporated information from one of the three sources attached to that event. Seeing the AI reference a supporting source was encouraging because I think that's where this feature has tremendous potential. Ancestry AI Fill in Timeline Gaps Feature- Example 3 A bit hit-or-miss (for now). After trying the feature on several facts, my overall impression is that the results are currently somewhat inconsistent. Sometimes the AI stays focused exactly where it should - on the individual timeline fact. Other times it drifts into summarizing the ancestor's life rather than describing the specific event. Fortunately, this isn't particularly surprising. The feature is clearly labeled BETA, and beta features exist so developers can identify issues, improve consistency, and refine the user experience before wider release. Where I think this feature could really shine. As someone who researches professionally, I can currently write these descriptions myself much faster and with greater precision. For me, it's still quicker to manually add information to a fact description than to generate AI text and then edit it. However, experienced genealogists are not necessarily the primary audience for this feature. The majority of Ancestry users are family historians who may not realize they can - or should - add descriptions to timeline facts in the first place. For those users, this tool could become incredibly helpful. If future versions consistently: * stay focused on the individual timeline fact, * summarize only the evidence related to that event, * incorporate all of the sources attached to that fact, and * generate accurate, well-written descriptions, then I think this could become one of Ancestry's more useful AI additions. My overall thoughts. I always enjoy seeing genealogy companies explore new ways to use artificial intelligence, and I appreciate that Ancestry continues experimenting with features that help users better understand and present their family history. Like many beta features, Fill in Timeline Gaps shows promise but still has a few rough edges. At the moment, I'd describe it as somewhat hit-or-miss, depending on the ancestor and timeline fact you're working with. When it stays focused on the specific event, it does a nice job of expanding the information. When it drifts into writing a mini biography for a single fact, however, it becomes much less useful. I'm optimistic that Ancestry will continue refining the feature as more people test it and provide feedback. As with any AI-generated content, remember that it should always be reviewed before accepting it. AI can be an excellent assistant, but it shouldn't replace your own analysis or careful evaluation of the records and sources supporting your research. For now, I'll continue experimenting with Fill in Timeline Gaps, and I'm interested to see how it evolves over the coming months. Have you noticed Ancestry's new AI "Fill in Timeline Gaps" feature yet? If you've tried it, I'd love to hear what you think. Did it stay focused on the individual facts, or did it wander into life summaries like some of my tests? Let me know in the comments! More Ancestry and Genealogy Resources. Learn about more Ancestry updates, tips, tools, and features and other family history and genealogy resources under the Ancestry.com and the Genealogy Resources categories of my blog and on my dedicated Genealogy Resources page. More AI tips, tools, and features. Learn about more AI in Genealogy tips, tools, features, and resources under the AI in Genealogy category.

AI + Genealogy monthly review - December 2025. Dec 31, 2025 December felt like a "new interface" month for family history: less search box only, more AI-guided storytelling, tree-building nudges, and machine-reading at scale. In this issue. * Ancestry turns records into narrated "AI Stories" * FamilySearch adds tree-extending hints powered by its AI Research Assistant * Genealogy education keeps leveling up (plus an Internet Archive + AI workflow) * Manuscripts meet machine learning: Cairo Geniza / MiDRASH momentum 1) Ancestry launches AI Stories (narrated audio from records). Ancestry introduced AI Stories in beta, generating a short narrated story from a historical record. Ancestry said 940M+ records are currently eligible for audio narration, and the feature supports multiple languages. Ancestry+1 Why it matters: this is AI being used as a presentation layer - helping families understand records faster, and making discoveries easier to share with relatives who won't read a census page for fun. 2) FamilySearch adds tree-extending hints via its AI Research Assistant. FamilySearch announced new home-page hints designed to surface record hints that could add a new spouse, parent, or child - with a workflow that still routes you back through review + source linking. familysearch.org Why it matters: it's a shift from "here's a record" to "here's what this record might unlock," which can reduce the time you spend staring at gaps in your tree. 3) Canada's 1931 Census index shows what "AI at scale" looks like in archives. Library and Archives Canada described the rollout of a free, searchable index for the 1931 Census of Canada (built in collaboration with Ancestry and FamilySearch) as part of its phased accessibility plan. Canada+1 Why it matters: beyond consumer genealogy sites, large institutions are now routinely using OCR/handwriting recognition + structured extraction to turn massive collections into searchable data. 4) the "machine finally learned to read" conversation goes mainstream (genealogy edition). A popular December post from AI Genealogy Insights argues that newer multimodal models (discussed there as "Gemini 3") are crossing a threshold for reading messy, real-world documents - while still requiring careful verification. AI Genealogy Insights Why it matters: for genealogists, the practical question isn't "is it perfect?" - it's "is it good enough to speed up my workflow without adding new errors?" 5) A smarter way to use AI: decompose first, prompt second. Another AI Genealogy Insights post offered a framework for breaking big research questions into smaller, testable tasks - so you can validate outputs step-by-step instead of trusting one giant answer. AI Genealogy Insights Try it: take one hard problem (e.g., "Who were John Smith's parents?") and turn it into mini-tasks: extract facts from each record | build a timeline | compare conflicts | propose hypotheses | list what evidence would confirm each. 6) Genealogy education keeps professionalizing around AI. Legacy Family Tree Webinars highlighted practical "best uses" of AI for genealogy - document understanding, extracting details, and turning research into readable narratives - plus a look at what to watch next year. Legacy Family Tree Webinars Why it matters: the community is moving from novelty prompts to repeatable, teachable methodology. 7) Internet Archive + AI: turning big books into working indexes. A webinar-focused deep dive showcased how AI can help build a surname index from a long digitized genealogy book and extract structured data (like migration routes) into spreadsheet-ready output. Legacy Family Tree Webinars Why it matters: this is "DIY indexing" - the kind of work that used to take days now becomes an afternoon project (with verification!). 8) medieval manuscripts meet deep learning: Cairo Geniza + MiDRASH momentum. Coverage this month spotlighted AI-assisted transcription and reconstruction of Cairo Geniza fragments (hundreds of thousands of manuscript pieces), including the MiDRASH effort and related "transcribe-a-thon" work using specialized manuscript platforms. Reuters+2RNS+2 Why it matters: it's a reminder that genealogy-adjacent breakthroughs often come from the wider archives/manuscripts world - and the tooling eventually trickles down into family history workflows. Key takeaways for December. * AI is becoming the "wrapper" around records (audio, summaries, guided prompts) - not just a search aid. Ancestry * Platforms are nudging users toward tree-building actions, while still keeping humans in the proof loop. familysearch.org * The biggest long-term unlock remains machine-reading at scale - especially for handwritten and semi-structured archival material. Canada+1 Ancestry AI Stories announcement (Dec 15, 2025): https://www.ancestry.com/corporate/blog/ancestry-brings-family-history-to-life-with-new-ai-powered-stori Semafor coverage (Dec 12, 2025): https://www.semafor.com/article/12/12/2025/ancestrys-new-ai-feature-narrates-ancestors-stories FamilySearch tree-extending hints (Dec 22, 2025): https://www.familysearch.org/en/blog/ai-research-assistant-home-page-hints LAC "1931 Census phase two complete" (Jul 30, 2025): https://www.canada.ca/en/library-archives/corporate/updates/2023/1931-census-phase-two-complete.html AI Genealogy Insights (Dec 16, 2025): https://aigenealogyinsights.com/2025/12/16/when-the-machine-finally-learned-to-read-gemini-3-and-the-question-of-good-enough/ AI Genealogy Insights framework (Dec 18, 2025): https://aigenealogyinsights.com/2025/12/18/the-art-of-breaking-things-apart-a-framework-for-ai-assisted-genealogical-research/ Legacy webinar: Best Uses of AI for Genealogists (Dec 19, 2025): https://familytreewebinars.com/webinar/the-best-uses-of-ai-for-genealogists/ Reuters on MiDRASH / Geniza (Nov 26, 2025): https://www.reuters.com/business/media-telecom/vast-trove-medieval-jewish-records-opened-up-by-ai-2025-11-26/ Religion News Service (Dec 17, 2025): https://religionnews.com/2025/12/17/ai-is-enlisted-to-rescue-jews-cultural-heritage-and-give-new-life-to-ancient-languages/ Times of Israel (Nov 25, 2025): https://www.timesofisrael.com/scholars-transcribe-hundreds-of-thousands-of-cairo-geniza-fragments-some-never-read-before/