I am part of the team that maintains FamilySearch GEDCOM 7 (hereinafter G7), but overall I would call GEDCOM X (hereinafter GX) a superior file format, though each has some strengths that the other lacks. However, GX has relatively little use as a file format (as of March 2026), leading to it being less practically useful as a means of data exchange.

This page tries to document significant differences between the two formats' conceptual models. It ignores the more superficial differences in syntax. Note that while I am deeply familiar with G7, I have more modest familiarity with GX so I might have missed something.

GX has two forms of relationship:

• gx:Relationship is a directed edge between two people.
• gx:Group is a conceptual collection of people, with gx:GroupRole indicating membership in the group.

G7 has more variation in how relationships are formed:

• g7:record-FAM is the primary relationship structure, encoding

  • Couple relationships with g7:FAM-HUSB and g7:FAM-WIFE

  • Parent-child relationships with g7:CHIL, including birth order information in the order of these and placeholders for children known to exist only by virtue of gaps in birth order.

    The meaning of a CHIL may be refined (and possibly even negated) by the INDI's FAMC.PEDI.

• g7:ASSO can also associate people with and g7:record-FAM, g7:record-INDI, or any fact or event.

When going from GX to G7, as the code in this repository does, I try to find relationships that look like FAMs and place them there, using ASSO for whatever is left. Going from G7 to GX would additionally have to decide when to make a gx:Group for a g7:record-FAM; I could see cases for never (FamilySearch doesn't seem to create gx:Group for each family made through it's UI), or only if the g7:recrd-FAM has an g7:ASSO (which otherwise would be hard to represent), or always (for consistency since it is sometimes needed for the g7:ASSO).

G7 has a structure g7:PHRASE that can appear in various places to allow users to enter unstructured text instead of structured information. GX does not have a similar structure. For GX-to-G7, this is not a concern: we can create g7:PHRASE when some part of GX has no G7 parallel and ignore it otherwise. For G7-to-GX, PHRASE would likely need to be converted (lossily) into something like a gx:Note.

Both G7 and GX allow notes in many places. There are some differences:

• GX allows notes in more places than G7.
• G7 has a distinction between inline notes (g7:NOTE) and linked, shared notes (g7:SNOTE). Converting between the two loses no information, but does lose that the information is linked which may change the semantics of future updates to the notes.
• G7 allows notes to be in several formats using g7:MIME; GX lacks this, but G7 includes guidance on converting to plain text.

G7 has several ways to record information that is not true:

• g7:NO expresses that something did not occur or was not true.
• g7:FAMC-STAT has several levels, including asserting that someone was not a child of someone else.
• g7:QUAY can express doubt about many different elements of the record.

GX has only one:

• gx:Conclusion's confidence fieldroughly matches g7:QUAY in that it can only express doubt, not a confident assertion that something is false.

For GX-to-G7, this is not a concern: we simply don't create g7:NO or g7:FAMC-STAT structures. For G7-to-GX, g7:NO and g7:FAMC-STAT would likely need to be approximated with something like a gx:Note. Additionally, a g7:FAMC-STAT value of g7:enum-DISPROVEN means there is not a gx:ParentChild relationship, and thus that there should be no gx:Relationship, which also makes it unclear where to place a gx:Note.

G7 personal names are somewhat ideosyncratic, storing the name string (g7:INDI-NAME) and the parts of the name (g7:GIVN, g7:SURN, and so on) with no requirement that the two be kept in sync. In many cases the name parts can be matched unambiguously to substrings of the name string to create a gx:NameForm with gx:NameParts inside, but in some cases this won't work. A common reason it might not work is the common practice of including alternative forms of name parts not used by the person but useful for searching and filtering, such as other-gendered forms of surnames.

GX names have a level of distinction that G7 lacks. Each gx:Name may have multiple gx:NameForm, allowing a distinction between unrelated names of a person (like "Augusta Ada Byron" and "Ada Lovelace") and different forms of the same name (like "Augusta Ada King, Countess of Lovelace" and "Ada Lovelace"). G7 has no similar distinction; names and name forms are both stored in g7:INDI-NAME with no way to indicate whether two g7:INDI-NAMEs represent forms of the same name or distinct names.

G7 has a structure g7:ALIA that allows asserting that two g7:record-INDI represent the same historical person. This structure is problematic even within the G7 ecosystem, with some applications using it extensively and others having no support for it at all. Those that do support it use it in different ways: some have a persona tree, others unmerged sets of linked personas, and I've heard rumor (but not yet see the files to support it) that some split a person along non-persona lines, for example storing "public" information in one INDI and "private" information in another with ALIA linkage.

GX has an gx:EvidenceReference which correpsonds one use of g7:ALIA: assembling an aggregate conclusion out of a tree of personas. Because of the spotty support for g7:ALIA in G7 tooling and the way different tools use g7:ALIA for different purposes, I've chosen not to impore gx:EvidenceReference to G7 at all. A G7 to GX converter would need to be cautious in confirming that g7:ALIA are used in a gx:EvidenceReference-compatible way (for example, without cycles, and ideally in an evidentiary way) before convering g7:ALIA to gx:EvidenceReference.

GX allows gx:Attribution in many places. G7 allows g7:SUBM and g7:CHAN much more sparsely, and with less scope for storing information about the change.

GX has a much richer set of event and fact types than G7 does event and attribute types.

G7 has more ability to refine event and attribute with g7:TYPE and with the generic g7:EVEN and g7:TYPE structures.

GX has a robust and unambiguous date format for the prolypic Gregorian calendar. G7 allows dates to be stored in multiple calendar systems.

G7 supports bounded approximate dates with BET/AND as distinct from multi-day events with FROM/TO. GX has just one form of data range; it is not clear to me from the specification whether that form is intended to be what G7 calls date periods (like WWII going FROM 1 SEP 1939 TO 2 SEP 1945) or date ranges (like the Siddhartha Gautama being born BET 563 BCE AND 400 BCE).