Ancestry.com, the world's largest online family history platform, has cut a record digitisation process that once took nine months of manual indexing down to under nine days, using proprietary handwriting-recognition AI to help push its collection past 71 billion records and restore subscription revenue growth after years of stagnation.
The US-based company, which operates in more than 30 countries and gives subscribers access to census, birth, marriage, death, immigration and military records alongside DNA testing and family tree building tools, has become one of the largest repositories of digitised historical records in the world. It was acquired by private equity firm Blackstone for US$4.7 billion in 2020.
Ancestry CTO Sriram Thiagarajan gave Authority Magazine a detailed account of the pipeline in June. Before 2021, records were scanned and then manually indexed through outsourced operations, a process taking almost nine months at roughly ten times the cost. Proprietary AI handwriting recognition compressed that to under nine days, and Ancestry has more than tripled its rate of content growth year over year.
"At Ancestry, we see AI as an amplifier of human capability, not a replacement," Thiagarajan said. "In our industry where we lift insights from historical records and validate deeply personal family histories, real human involvement is critical." He said human review remains embedded in the workflow to guard against AI hallucination and to preserve historical accuracy and cultural sensitivity.
The turnaround has also been detailed by Ancestry's CEO. Howard Hochhauser told Semafor's CEO Signal in April that the company plans to invest US$450 million over the next 10 to 15 years in digitising new records and applying AI to extract more accurate information from previously scanned material.
Hochhauser, a former CFO who became CEO in early 2025, told Semafor the company had drifted from its core customer under a strategy of making the site more social. "Revenues flattened out, customer satisfaction declined," he said. Refocusing on content and storytelling has since delivered five consecutive quarters of subscription revenue growth and 13 months of rising customer satisfaction scores.
The content engine is archival. Hochhauser described photographing paper voting registers in the basement of Manhattan's City Hall, after which "technology takes over and AI will take that handwritten script and make it searchable for everybody. Every day, we're writing 10 million new records to the site." The company is investing in birth, marriage and death records in the US, UK, Canada and Australia, with Italy and Poland next.
The generative layer sits on top of the structured extraction. An AI Stories feature, launched in December, turns static documents such as census records into narrative and audio stories; more than 500,000 unique users have engaged with it and almost a million audio stories have been played. AI translation has also enabled rapid international expansion, with the site, content and marketing campaigns launched in Italy within a month.
Technology publication Let's Data Science, analysing the interviews, called Ancestry a case study in how domain-specific OCR, generative storytelling and human-in-the-loop validation can compound into a defensible data advantage over a decade. It flagged error rates on non-English historical scripts, dataset provenance and cross-border privacy obligations as the issues to watch as the approach scales internationally.
Hochhauser also noted the limits customers place on automation. "It's somebody's hobby, the joy is in the doing and discovery," he told Semafor. "You don't want your AI assistant to do your crossword puzzle for you; that'd be wholly uninspiring."