Worked examples
These cases are real-shaped but entirely fictional — no real person, place, fond number or DNA result appears anywhere on this site. They show how the same evidence rules produce a confident conclusion, an open conflict, or a clean rejection.
Anti-ThruLines
A predicted-relative line drawn through named ancestors reads like a finding, not a guess. It is built from other people's trees, it looks authoritative, and it silently reshapes how a branch appears to connect. That is structurally risky — and a single prediction can stack several errors at once.
Here is the canonical three-error stack, where one ~25 cM side-cousin is rendered as a direct grandmother:
01
A distant cousin is placed on the maternal collateral line, not the line where the archival record actually puts the person. The prediction chose the branch that made its line shortest — not the one the evidence supports.
02
An intermediate ancestor is compressed by a generation. Collapsing one generation shortens the predicted path and lets a distant collateral masquerade as close family.
03
The result is labelled a direct grandmother — but a true grandmother shares roughly 1,700 cM, and this match is around 25 cM. The cM value alone falsifies the placement. The label and the DNA cannot both be true.
Each error is survivable alone. Stacked, they convert a side cousin into a fake direct ancestor that then anchors more fantasy above it. AutoTreeGen compares the cM value against the expected band for the claimed relationship, looks for a primary record bridging the generations, and checks for a better-supported placement. When the checks fire, the placement is rejected — never silently applied.
How the engine records it
Assertion
Evidence (3)
Popularity is not evidence; the chain has no primary source behind it.
The suggested birthplace is anachronistic for the recorded year.
Expected a match at this distance; none observed across three kits.
The same evidence discipline applies to every claim. The difference is only what the records and DNA actually support.
Assertion
Evidence (3)
Autosomal match consistent with a close-cousin range on the paternal line; single long segment, not yet triangulated.
Birth registration naming Johann Keller as father; transcription confidence high.
Vital-record year and DNA-implied generation gap disagree; unresolved.
Assertion
Evidence (3)
Maternal name recorded at registration; legible, signed by the registrar.
Mother and daughter listed in one household; ages consistent with the birth register.
Two maternal-line descendants share the same mitochondrial haplogroup; consistent with a direct maternal line.
Assertion
Evidence (2)
Surname pattern suggests a Kohanim line; in an endogamous community a shared surname is a hint, never an attribution.
A famous-line carry requires Y-DNA confirmation. Without it — and given deep endogamy — this stays a hypothesis.
Assertion
Evidence (3)
Primary registration entry, district archive, fond 503.
A second primary source names a different father for the same daughter.
Match favours the Wojciech line, but does not by itself resolve the document conflict.
Assertion
Evidence (3)
Popularity is not evidence; the chain has no primary source behind it.
The suggested birthplace is anachronistic for the recorded year.
Expected a match at this distance; none observed across three kits.