A misread name is the one mistake that can't happen. ArchiveLM flags what it can't read instead of guessing.
Locating heirs means reading handwritten wills, parish registers, court records and census pages, often in other languages, line by line. It's the slowest part of every case, and a misread name means the wrong heir.
Every record mentioning a Juan or John Herrera born around 1880
Baptisms in this parish for the Novak family
Wills naming the property on Mill Road
Load scans of wills, parish registers, court records and census pages for the case you're working.
AI reads 18th- to 20th-century handwriting in any language, and refuses a page it can't read reliably rather than guessing a name.
Records are indexed by meaning, with names, dates and places captured — including common name variants.
Search a name across every record it appears in, in plain language, with every finding cited to the original page.
Reads 18th to 20th-century handwriting, in any language
Finds name variants across records (Juan / John, Novák / Novak)
Refuses an unreadable page rather than guessing, so a wrong name never enters your file
Every finding cited to the original page for your report
Each account's documents sit in an isolated database and storage bucket, enforced by row-level security — no other customer can see or search them.
Encrypted in transit (TLS 1.2+) and encrypted at rest.
Your documents are never used to train AI models — we instruct our providers not to train on your data.
Documents are stored in the United States.
Deleted on request, and removed within 30 days of an account closing.
A published list of every subprocessor that touches your data, with its role and location (legal.archivelm.com/legal/subprocessors on the main site).
We are not yet SOC 2 certified. We'll share our security overview and subprocessor list on request.
| Plan | What's included | Price |
|---|---|---|
| Try it | ||
| Explorer (Free) | 25 credits/month — enough to load your own hardest pages and see what it finds | $0 |
| Plans | ||
| Researcher | Individual plan — 100 credits/month, semantic search, AI Librarian chat, Research Lab, unlimited exports | $15/mo |
| Team | Up to 5 seats — 420 credits/month, all research features | $49/mo |
| Department | Up to 15 seats — 2500 credits/month, priority support | $249/mo |
| Large backlogs | ||
| Project pricing | A one-off bulk digitization project for a large backlog, priced per page instead of by plan | from 6¢/page at volume |
A dense newspaper page uses more credits than a book or letter page, because it takes more work to read. Ask us which plan fits your backlog before you commit.
You bring the documents — wills, parish registers, court records and census pages you already hold or can obtain. ArchiveLM doesn't supply source records; it reads, indexes, and makes searchable what you load into your own account.
It's flagged, not guessed. Historical scans degrade, fade, and get handwritten over — when confidence is too low to trust, the page is marked for review instead of silently producing a confident-looking answer that isn't on the page.
Yes. Semantic search works across languages — an English query can surface a baptism or marriage record kept in Latin, Spanish or Polish by meaning, not just keyword match, and the AI Librarian can answer in English while citing the source-language passage it relied on.
Row-level security isolates every account's documents in their own database rows and storage bucket. No other customer can see or search your collection, and nothing you upload is used to train AI models.
The free tier includes 25 credits a month, enough to load a handful of your hardest pages and see exactly what it finds and where it fails. Paid plans start at $15/month; large one-off backlogs are priced per page instead.
Yes — searching for a name surfaces common variants and transliterations across records, which matters when the same person appears as "Juan" in a baptism record and "John" in a probate filing decades later.
Send up to 50 pages from a real or closed case — the handwritten, faded or foreign-language ones. We'll run them and show you exactly what ArchiveLM finds, and where it can't read a page.
ArchiveLM is in private beta. We review each request and typically respond within 1–3 business days.