Volunteer-indexed genealogy records are full of transcription errors that make ancestors unsearchable, and corrections take months to propagate

FamilySearch, the largest free genealogy platform, relies on volunteer indexers to transcribe handwritten historical records into searchable text. When a volunteer mistranscribes a name — reading "Kaczmarek" as "Raczmarek," or "Margarethe" as "Maryarette," or interpreting old German Kurrent script "sch" as "ſch" — that ancestor becomes effectively invisible in search results. The original document image exists on the platform, but if you do not already know the exact record to look at, the mistranscribed index entry means your search will never find it. FamilySearch acknowledged this problem and in 2019 began allowing users to submit corrections to name fields, later expanding to dates and places. But corrections do not replace the original transcription — they add an alternative that coexists with the error, and the correction process can take anywhere from minutes to months to become searchable. This is devastating for researchers working with non-English records. A Polish parish register written in Latin with German administrative notes and Polish place names can confuse even experienced indexers, let alone volunteers with no language training. The same ancestor might be recorded as "Wojciech" in one document, "Adalbertus" (the Latin equivalent) in another, and "Albert" in a third — and if the indexer transcribes any of these incorrectly, the link between records is broken. For immigrants to the US whose names were anglicized at various points, the compounding of original-language transcription errors with anglicization variants creates a search space so large that many ancestors are simply never found. The structural cause is that volunteer indexing was designed for scale, not accuracy. FamilySearch has indexed over 8 billion names using millions of volunteers, and the quality control mechanism — having multiple volunteers independently index the same record and comparing results — still cannot catch errors when all volunteers lack the language skills needed for a particular record set. There is no requirement that indexers speak the language of the records they transcribe. OCR technology cannot reliably read pre-20th-century handwriting in any language. And the correction system, while a step forward, shifts the burden to the end user who must already know what the correct transcription should be — which defeats the purpose of an index.