You want to leave. Maybe the subscription isn't worth it anymore, maybe your research has moved to a region another platform covers better, maybe you're just tired of the interface.
And the thing stopping you is a reasonable fear: that twenty years of work is going to arrive at the other end as a pile of names with everything useful stripped out.
That fear is about sixty percent justified. Here's the accurate version.
The thing to understand first
You are not moving your tree. You are moving the structure of your tree.
Every platform-to-platform move runs through a GEDCOM file — a plain text format that's been the genealogy interchange standard for decades. It carries the skeleton: who existed, how they were related, names, dates, places.
It does not carry the flesh. The record images, the photographs, the platform's own hints and matches — none of that is in the file, because none of it belongs to you in a transferable sense. It belongs to the platform's record collection, and you were renting access.
So the accurate mental model is: you export a skeleton, import it somewhere new, and then rebuild the flesh from the new platform's records. Which is work — but it's also frequently productive work, because the new platform surfaces documents the old one never had. More on what's actually inside a GEDCOM.
What survives, exactly
| Transfers | |
|---|---|
| People and relationships | Yes, reliably |
| Names, including alternates | Yes |
| Birth, marriage, death dates | Yes |
| Places | Yes |
| Source citations (the text) | Mostly |
| Notes | Usually, sometimes truncated |
| Custom fields and tags | Often not |
| Attached record images | No |
| Photographs | No |
| DNA matches | No |
| Hints and suggested records | No |
| Your research log and to-dos | Usually not |
Three of those need explaining.
Source citations transfer as text, not as documents. The citation saying a birth date came from a particular census page comes across. The census page does not. You arrive with a footnote pointing at a document you can no longer see, unless you saved a copy.
Photographs are the one that surprises everyone. GEDCOM stores file paths to images, not the images themselves. Those paths point at locations on the source system, which means they resolve to nothing at the destination. Your photos are not in the file. They have to be moved separately, by hand, and re-attached.
Custom fields and program-specific tags degrade. Every program adds its own extensions for things the standard never covered. Those get dropped in transit, silently. If you've built a workflow around a field unique to one program, assume it won't survive.
The sequence
Order matters, and getting it wrong is how people actually lose work.
1. Export before you cancel anything. Obvious, universally ignored. The export function lives behind the subscription on most platforms. Cancel first and you may find the door locked from the outside.
2. Save your record images and photographs first, separately, before the export. This is the step that determines whether the move costs you the documentation. Every record image you attached is retrievable while your subscription is live and gone when it isn't. Download them to your own storage, named so you can match them back to people. Tedious for a large tree, and it is the difference between moving your research and moving an index of it. Naming conventions that work.
3. Keep the original export file untouched. Archive a pristine copy before you import it anywhere. Some programs alter a file on import. That file may be the only complete record of your work at a moment when it exists in exactly one place.
4. Import, then verify counts on both sides. How many individuals did the old platform say you had? How many does the new one show? A discrepancy means something didn't come across, and you want to know that immediately rather than in four months.
5. Spot-check your hardest ancestor. Pick the person whose record is most complicated — multiple marriages, conflicting dates, a disputed parentage. If that person came through intact, the simple ones did too.
6. Only then cancel.
For the actual export mechanics — which menu, which screen, which format options — those differ enough per platform that reproducing them here would go stale within a release. The per-app export guides at brightbranches.com cover Ancestry, Family Tree Maker, RootsMagic, Legacy, and MyHeritage individually and are kept current.
The version-number thing
Two GEDCOM versions are in circulation. 5.5.1 has been the working standard for decades and is supported essentially everywhere. Version 7.0 is newer and support is uneven.
Practical advice: export as 5.5.1 unless you have a specific reason not to. The newer version handles some things better, but a file the destination can't fully read is worse than a file in an older format that everything understands.
When the move isn't a move
Two situations where the mental model above doesn't apply.
Moving into FamilySearch's shared tree is not an import. That tree has one profile per deceased person, contributed to by everyone. Adding your research means reconciling it person by person against profiles that may already exist — a real project, not a button. Worth doing, and not something to attempt on a deadline. How that model works.
Moving to desktop software is a different proposition entirely, and often a better one. Your data lives on your machine, in a file you control, and it stops being dependent on anyone's subscription or corporate continuity. GEDCOM import into desktop programs is generally cleaner than platform-to-platform, because the desktop programs have been reading these files for thirty years.
The habit this should leave you with
Whether or not you actually switch, this exercise reveals something worth acting on: your research is more dependent on one company than you assumed.
The fix is small and permanent. Export a GEDCOM once a year and keep it somewhere that isn't a company. Save record images as you attach them rather than relying on the attachment. Keep your photographs in folders you control.
Do that and switching platforms stops being a migration project and becomes a preference. It also means your research survives things much worse than a platform change. More on that.
Keep reading: