AutoTreeGen — Evidence-First Genealogy Intelligence
A research instrument

The library
of your
lineage.

AutoTreeGen is a working archive for serious genealogy. Every name, every date, every DNA segment is bound to a citation, a confidence score, and a chain of provenance you can defend in print.

250Kpersons per tree on Pro
8audit engines
QUAY 1–3citation grades
46languages in the app
The thesis

Genealogy is not a hobby. It is the guesswork evidence we owe to the people who came before us — and the ones who come after.

No. 01

Cite, or it didn't happen.

Every fact traces back to a source — a parish register, a census line, a centiMorgan segment. Unsupported claims are stored separately, as hypotheses.

QUAY 1 · 2 · 3
No. 02

Disagreements are versioned, never overwritten.

Two cousins, two birth years, two scribes' hands. The archive keeps all of them. Confidence is computed, not asserted, and updates with new evidence.

audit trail · permanent
No. 03

Paper and DNA, in one workspace.

Y-DNA, mtDNA, autosomal cM segments and your scanned records sit in the same tree. The engine reads both at once.

Any GEDCOM · strict
The archive

An archive that thinks in centuries.

Fig. 01

Eight lineages, four generations, one diagram.

The signature ribbon-tree: each branch carries its lineage colour from root to tip, with portraits attaching where photographs survive.

Fig. 02

Confidence as a number, not a feeling.

Support and contradiction are weighed apart, each piece by the grade of its source, and the report writes down the arithmetic the number came from.

Fig. 04

QUAY-graded provenance.

Sources are graded on a 1–3 scale, and the grade travels with the claim it supports.

Fig. 05

Y, mt, and autosomal in one view.

Three evidence streams — paternal line, maternal line, recent cousins — overlay on the same tree.

Do not build trees of names. Build trees of evidence.
— the founder

No fact without a source.

Every name, date and DNA segment is bound to a citation, a confidence score and a provenance trail you can audit — and defend.

See it work

Eight engines, one workspace.

Pick an engine to watch it run on a sample tree — the names and places are synthetic.

GEDCOM Doctor
# GEDCOM Doctor · read family.ged✓ parsed 41,208 persons · UTF-16 → UTF-8▲ 37 broken FAMS links · 12 duplicate persons✓ merged Müller/Mueller · parents + sources keptrepaired · nothing lost · reversible
Claim under auditRejected

Helga Vogt is recorded as the mother of Klara Müller.

autosomal · 0 cM shared, ~3,400 expected mother–child−0.62
autosomal · 1,540 cM across a known descendant pair+0.30
baptism register 1849 · names Margarethe Weiss+0.80

Timeline impossible and DNA excludes maternity. Claim removed from the canonical tree, kept in the audit trail. Suggested correction — mother is Margarethe Weiss.

Evidence graph
Klara MüllerMargarethe WeissHelga VogtHans MüllerL. Kovač?DNA
ConfirmedProbableHypothesisConflictDNA support
Working in the background

Your evidence, being assembled.

DNA kitsComputing
Hypotheses trackedComputing
Conflicts openComputing
Sources gradedComputing
Methodology

From GEDCOM and DNA to a defensible tree.

i

Import

Drop any GEDCOM file — from Ancestry, MyHeritage, Geni, any version, any size. We parse it strictly and tell you which records bend the spec.

Any GEDCOM · any size
ii

Match

Add Y, mt and autosomal kits, or your exports from Ancestry, MyHeritage and GEDmatch. Each kit becomes its own evidence stream.

cM · segments · haplogroups
iii

Hypothesise

The engine proposes parent–child links, scores them, and records the evidence both for and against.

confidence ∈ [0, 1]
iv

Cite

Every finding carries its sources, each with its own grade.

grade shown on every source
Why AutoTreeGen

What the engine actually does.

It checks the evidence

Every proposed link is scored against the records and DNA behind it; anti-fantasy detectors flag the fabricated and the over-reached.

It keeps the chain of sources

Each fact stays bound to its citation — parish register, census line, DNA kit — graded QUAY 1 to 3.

It accounts for DNA

Y-DNA, mtDNA and autosomal cM read alongside the paper trail, with endogamy weighed in so shared segments are never over-read.

It tracks conflicting hypotheses

When sources disagree, both are kept and versioned. Confidence is computed from the evidence, never overwritten.

It writes a proof report per relationship

One claim at a time: the conclusion, the evidence for and against it, footnotes in Chicago style, and a certificate that fixes the file and the rules it was made by.

What is built

Already working, today.

Several trees become one

Preview the merge, resolve the conflicts, commit into a new tree. Nothing is lost and every decision is kept.

Duplicates found and merged in bulk

The Duplicate Detective proposes pairs, and you accept them one by one or hundreds in a single action.

A conflicts queue where a person decides

Contradictions between sources wait in a queue, and a person closes each one: confirm, reject, keep it as a hypothesis, or override it with a reason on the record.

One door for every DNA file

Ancestry, MyHeritage, FamilyTreeDNA, 23andMe, GEDmatch, AutoCluster ZIP — recognised by content, never by file name.

Kits from different sites recognise the same person

A match seen on two platforms is proposed as one person — and nobody is merged without you.

Haplogroups on the tree

The Y and mt lines read from your kits become a claim on the paternal or maternal line, shown where that line runs.

Branch Curator

Every branch outside the main tree, listed: keep it, attach it or delete it — and undo if you change your mind.

Your file, signed

Every GEDCOM the platform writes is identifiable and verifiable, and your original stays downloadable at any moment.

46 languages, and surnames found by sound

The app speaks 46 languages, and Daitch–Mokotoff phonetic search finds a surname however the clerk spelled it.

A feedback button in the top bar

Say what broke and paste a screenshot straight from the clipboard — it reaches the founder together with the page it was sent from.

Scale

Built for the trees that break other tools.

Serious research outgrows casual software. AutoTreeGen's import pipeline is tested end-to-end at 350,000 persons — a real parse, real persistence — on demand, on our own hardware. Free starts at 5,000 persons, Pro carries 250,000, Premium is unlimited.

Trust

Your DNA is special-category data. We treat it that way.

Encrypted at rest. Never used to train models. Every DNA computation runs behind explicit consent — no consent, no processing, engine by engine. Deletion is cryptographic: wipe the key and the data is gone. Your segments feed your tree's evidence and nothing else.

Uniqueness

Provenance everywhere. No one else builds a tree this way.

A fact here is an assertion, its evidence, and a computed confidence — never a bare name. Conflicts are versioned, not overwritten. The Anti-Fantasy detector rejects the impossible. Our mtDNA reference carries all 5,371 haplogroups of PhyloTree Build 17 in van Oven's exact notation — and when a chip cannot distinguish subclades, we say so instead of guessing. Many of these instruments exist because researchers on genealogy forums around the world asked for them — endogamy-aware DNA, honest chip resolution, imports that survive a million persons. Built to requests, not to trends.

Support

When it matters, a person answers.

  • Community · Free
  • Email 12h · Pro
  • Dedicated · Premium
Pricing

Plans built around evidence, not vanity

Three plans — Free, Pro and Premium. Free is a real working tier: a tree, 5,000 people, the GEDCOM Doctor and the Anti-Fantasy detector, no card required. Pro and Premium add full DNA and the advanced engines. Capability is the contract.

2 months free
Carved oak medallion — Free plan
Free
Carved oak
Open a tree, read the record, see what’s there.
$0 USD forever
Start free
Polished silver medallion — Pro plan
Pro
Polished silver
Merge matches, resolve conflicts, grow the tree.
$290 USD/yr
2 months free
Get Pro
FAQ

Questions, answered plainly.

What file formats can I import?

Any GEDCOM — from Ancestry, MyHeritage, Geni or any other tool, any version. We parse it strictly and flag the records that bend the spec.

Is my DNA data safe?

DNA is special-category data: encrypted at rest, never used to train models, and processed only under your explicit consent. You can delete it at any time.

How is confidence calculated?

Every claim is graded on its evidence — sources scored QUAY 1 to 3, DNA weighed against the expected range, agreement between records. Confidence is computed, not asserted.

Can I export my tree?

Yes. Your tree exports back to GEDCOM, and any relationship inside it can be argued in a PDF proof report of its own.

Is it available now?

Yes. AutoTreeGen is open: create an account and import your GEDCOM. There is a feedback button in the top bar — tell us what breaks.

Get started

See the proof behind your own tree.

Create a workspace and upload your GEDCOM. The engine grades every claim with a status, a confidence, and the evidence behind it.

DNA is special-category data — encrypted at rest, never used to train models, handled under explicit consent. The details live in the privacy notice.