Why Your Ethnicity Estimate Changed: Reference Panel Updates Explained

If your AncestryDNA breakdown shifted overnight from 32% Scottish to 41% Irish, your DNA did not change. The company refreshed its reference panel and recomputed every customer’s estimate against it. Reference panel updates are one of the least intuitive parts of consumer DNA testing, and they happen often enough that customers blame the test for being unreliable when the test is doing exactly what it was designed to do.

What a reference panel update actually is

Every ethnicity estimate is a comparison against a reference panel: a curated set of people the company has identified as having four grandparents from a single region. When a company adds more reference samples, removes mislabeled ones, redraws region boundaries, or improves the algorithm, every customer’s breakdown changes the next time it is recomputed.

This is not a bug. A larger panel with more carefully labeled samples produces a more accurate breakdown. The cost is that the breakdown looks different than it did before, sometimes substantially.

A timeline of major updates

AncestryDNA has refreshed its ethnicity estimate roughly every other year. The 2018 update expanded sub-regional coverage of the British Isles and Scandinavia. The 2020 update added significant new African and Indigenous American sub-regions, which shifted many customers’ breakdowns in underrepresented categories. The 2022 update introduced “Communities” as a distinct overlay and refined Eastern European sub-regions.

23andMe has updated its Ancestry Composition module roughly twice a year on a rolling basis. Customers usually see a notification in their account when a refresh is applied. MyHeritage released a v2.0 ethnicity estimate in 2019 that was a substantial rewrite. The companies do not always announce minor updates publicly, which is part of why a long-time customer can log in one morning to a noticeably different breakdown without warning.

What kinds of changes to expect

Continental-level percentages usually move a few points or stay roughly flat. Sub-regional percentages can move much more, especially in regions that share recent ancestry with their neighbors. A breakdown that was “42% England, 28% Scotland, 12% Ireland” can become “31% England, 38% Scotland, 11% Ireland” after a UK-focused panel expansion, because the algorithm is now better at telling those neighbors apart.

Estimates can also reverse direction. A small Scandinavian percentage that turns out to have been Northern European drift can disappear in a refresh. A small Iberian percentage can appear if the panel finally has enough Iberian samples to detect a real signal that the older panel missed.

What it does not mean

It does not mean the test was wrong before, or that it is right now in some absolute sense. Both estimates are the company’s best guess against the panel it had at the time. The new estimate is usually better, in the sense of being a closer match to the real distribution of your ancestry across the company’s region definitions, but “better” is statistical, not absolute.

It also does not mean your relatives’ estimates will change the same way. Two siblings inherit different sets of DNA from the same parents, and a panel update can move their breakdowns in slightly different directions. This is normal and expected; we cover the underlying mechanism in understanding ethnicity estimates.

What to do when your estimate changes

Read the company’s release notes for that update, if they published any. AncestryDNA, 23andMe, and MyHeritage have all written explainer posts for their major refreshes. Compare what region definitions changed against the shifts in your own breakdown.

If you have built a tree based on a particular region showing up, do not delete branches based on a panel refresh. The matching side of your result, the actual shared DNA with real people in the database, is far more informative for genealogy than the ethnicity estimate. Our how DNA matching works guide covers what those matches tell you, and ethnicity estimate accuracy covers what a percentage is and is not reliable for in the first place.

For the broader picture, see our ancestry and genealogy pillar guide.