KleioBase vs Ancestry: Records Platform or Research Layer?

Ancestry now reads handwriting in two distinct places, and conflating them is the easiest mistake to make when working out whether you still need anything else.

The first is full-text search across Ancestry's own collections, which turns previously unindexed record sets into searchable text. The second, newer and more relevant here, is Image Transcript: upload a handwritten document to an ancestor's gallery in your tree, and Ancestry's AI produces a transcript in a panel you can read, copy and edit. Ancestry has described itself as the global leader in handwriting recognition, and it has the census-scale track record to back the claim, having run recognition over collections such as the 1950 US federal census and the 1921 census of England and Wales.

The short answer. Ancestry is a records repository that has added AI reading on top. KleioBase is a research layer that reasons over whatever you have gathered, from any source, and holds no records of its own. If you are choosing one, you are almost certainly choosing wrong, because for most researchers these are not alternatives. The real question is not whose transcription is better. It is whether anything in your workflow currently tells you that two of your profiles are the same man, that three of your sources disagree about a birth year, or that a generation is missing from a line.

A transcript attached to an image is a filing improvement

Image Transcript does something genuinely useful. A letter in Kurrent that you could not read is now text you can read. That is not nothing, especially for correspondence, journals and the kind of family documents no repository will ever index.

But look closely at where the output lands. It is a transcript, attached to one image, in one ancestor's gallery. The document has been made legible. It has not been made analysable. If the letter names four people, three towns and two dates, all of that is now sitting in a text panel where you, personally, still have to notice it, decide who those people are, and attach them to the right profiles by hand.

Reception has not been uniformly positive either. At least one reviewer comparing it against other AI transcription tools found the accuracy lower than alternatives tested. As always, and this is advice that applies to us equally, the original image is the authority and a questionable reading should be checked against it.

Reading a document is an input, not the product

Worth stating plainly, because the shape of this comparison hides it: KleioBase is not a transcription tool. Transcription is one of two ways to get data in. The other is a GEDCOM import, and a researcher who imports a tree and never uploads a single scan still gets the whole product - duplicate detection across the tree, conflicting dates surfaced as conflicts rather than silently overwritten, missing generations flagged as gaps, a tree health score, and a companion that can be questioned about the whole archive in plain language.

That is the product. The reading step exists so that documents can join the same analysis, not because text is the deliverable.

Ancestry's real product is the records, and that is not a small thing

It is worth being direct about the asymmetry here. Ancestry holds collections you cannot get anywhere else, particularly for the United States, the United Kingdom, Canada and Australia. Its hint system surfaces records you would not have thought to look for. Its DNA database is among the largest in the world and connects you to living cousins.

None of that is something we do or intend to do. We have no records. We have no DNA. We have no hints drawn from an external index, because there is no external index here.

Anyone telling you to cancel a repository subscription and buy an evidence tool instead is selling you something badly. The comparison worth making is not repository against evidence layer. It is whether, once you have paid for the repository, the transcript it hands you finishes the job.

Where the work actually goes

Here is the shape of the problem a transcript does not touch.

You have gathered thirty documents about one branch. Six came from Ancestry, four from a regional archive in Poland, eight from a cousin's shoebox, five from FamilySearch, and the rest from correspondence with a parish. They are in three languages. The same woman appears in eleven of them, spelled six ways, because clerks spelled by ear and nobody asked her.

Every one of those documents is now transcribed. You are not finished. You are at the beginning of the part that takes months: deciding which mentions are one person, which are two, what her actual birth year was given four documents disagree, and which generation is missing.

That reconciliation is the work. It is also the work that no transcription feature, ours or anyone's, does by producing text.

Side by side

When Ancestry is the better answer

You need the records. This is most people most of the time. If your ancestors are in Ancestry's core geographies and you do not yet hold their documents, the subscription is the purchase that matters and nothing here changes that.

You want DNA matching. We do not do this at all.

You want hints. Ancestry surfaces records against your tree automatically from its own index. We cannot, because we have no index to surface from.

You want one platform. Tree, records, DNA and now transcription in one subscription is a real convenience, and splitting tools across vendors has a genuine cost in friction.

What KleioBase does instead, and what it does not

We are a tool in this space, so read this as disclosure rather than a neutral verdict.

KleioBase is built for the archive, not the page. Everything you add, whether it arrives as a GEDCOM or as a photograph of a parish register, becomes the same structured model of people, dates, places and relationships, and the analysis runs over all of it at once. Eleven mentions of a woman spelled six ways become one profile with a documented life and eleven sources behind it. Conflicting dates surface as conflicts rather than quietly overwriting each other. Likely duplicates are flagged for deliberate merging. Missing generations show up as gaps. And you can ask the whole archive a question in plain language instead of reopening thirty files.

Documents are read and translated on the way in when they need to be. That is a capability, not the proposition.

It is private, and it is portable: GEDCOM in and GEDCOM out, with the caveats that apply to any GEDCOM transfer.

Now the honest limits.

No records, no hints, no DNA. Three of Ancestry's four main reasons to exist are simply absent here. This is a deliberate scope choice, not a roadmap gap.

You still need a repository. Realistically you will pay for both. We are an addition to a research budget, not a substitution in it.

Everything is confirmed by you. Extraction is reviewed before it enters your archive. That is the right default for evidence, and it does mean a thirty-document backlog carries real review time.

Accuracy is not perfect and we will not claim it is. Faded or heavily stylised hands are hard for any model and for most humans. Check readings against the image, exactly as you should with Ancestry's.

Where to start on your next branch

Pick the ancestor whose file is thickest and most confusing, the one where you have opened two documents side by side more than once to work out whether they are the same person. Keep your Ancestry subscription and keep using it to find records. Then take that one branch, the whole messy folder of it including the documents that never came from Ancestry at all, and put it somewhere that reconciles the people rather than just rendering the text.

If the branch resolves into fewer people than you thought it had, that is the answer. If it does not, you have lost an afternoon and learned that your filing was better than you feared.