AncestryDNA identifies close relatives with near-perfect accuracy, but match reliability drops sharply as the relationship gets more distant. For parent-child and sibling connections, the system is essentially foolproof. For third cousins and beyond, a significant percentage of reported matches may not reflect a real recent common ancestor at all.
Understanding where the accuracy breaks down, and why, helps you figure out which matches in your list are worth pursuing and which ones are probably noise.
Close Relatives Are Almost Always Correct
AncestryDNA assigns parent-child and full sibling relationships with very high confidence. These close connections share so much DNA that there’s virtually no room for error. A parent and child share roughly 50% of their DNA, and full siblings share around 37% on average. These amounts are large enough that no statistical quirk or technical limitation can produce a false match at this level.
First cousins are also highly reliable. They share an average of about 866 centimorgans (cM), which is the unit used to measure shared DNA segments. The range runs from roughly 396 to 1,397 cM, but even at the low end, that’s a substantial amount of shared DNA that clearly signals a real biological connection. If AncestryDNA tells you someone is your first cousin, you can trust it.
Second and Third Cousins Get Murkier
Second cousins share an average of 229 cM, but the range stretches from as low as 41 cM to as high as 592 cM. That wide spread means a second cousin could share roughly the same amount of DNA as a first cousin once removed, or as little as a third cousin. The match itself is almost certainly real at this level, but the predicted relationship label may be off by a step or two.
Third cousins are where things start to thin out. They share an average of just 73 cM, with a range from essentially zero to 234 cM. That zero at the bottom is important: it means two people who genuinely are third cousins can share so little detectable DNA that the test doesn’t pick up the relationship at all. By the third cousin level, you’re only expected to detect the connection about 90% of the time. Some real third cousins simply won’t appear in your match list.
The Danger Zone Below 15 cM
The real accuracy problems show up with distant matches sharing small amounts of DNA. Genetic genealogist Blaine Bettinger analyzed his own match data and found a clear threshold: above 15 cM, a match has a 99.3% probability of being a genuine connection inherited from a shared ancestor.
Below that line, the numbers fall fast:
- Below 10 cM: Only 59% of matches are confirmed as real. That means 41% of your matches in this range don’t actually share DNA inherited from a common ancestor.
- Below 7 cM: Only 40% of matches are real. The majority, 60%, are false positives.
These false matches aren’t errors in the lab work. They happen because very small DNA segments can appear identical between two people purely by chance. Your DNA is a mosaic of common human genetic patterns, and when two unrelated people happen to carry the same short sequence, the algorithm reads it as a shared ancestor. The shorter the segment, the more likely this coincidence becomes.
AncestryDNA uses a filtering system called Timber that removes segments of shared DNA deemed “uninformative,” essentially tossing out stretches that are so common across the general population that they don’t indicate a real genealogical connection. This means AncestryDNA is already more conservative than some other testing services. Matches that would appear on other platforms sometimes don’t show up on Ancestry at all. Even with this filter, though, the small-segment problem persists for distant matches.
Why Endogamy Inflates Your Matches
If your ancestry traces back to a community that married within itself for generations, your match list is probably messier than average. This pattern, called endogamy, affects groups including Ashkenazi Jews, Low-German Mennonites, and many island or tribal populations. After centuries of relative genetic isolation, everyone in the group shares an unusually large number of small DNA segments from distant common ancestors.
The practical effect is that two people from the same endogamous group will share significantly more total DNA than their closest genealogical relationship would predict. Someone who is genuinely your third cousin might share DNA equivalent to a second cousin or even closer, because you’re also connected through dozens of other distant ancestral lines. You might be third cousins through one line, but also fifth cousins three different ways, sixth cousins twice over, and so on. All those tiny shared segments add up.
This inflated sharing throws off AncestryDNA’s relationship predictions. The system sees a large total and assumes a closer relationship than actually exists. One practical workaround is to mentally apply a higher threshold. Some genealogists working with endogamous populations exclude all segments below 7 cM from their calculations. An individual who appears to share 120 cM with you might only share 60 or 70 cM of genealogically meaningful DNA once those small, population-wide segments are stripped away.
What This Means for Your Match List
Your AncestryDNA match list is best understood as a gradient of confidence. At the top, close family matches are rock solid. In the middle, second and third cousin matches are almost certainly real connections, though the exact relationship label may be slightly off. The system might call someone a second cousin when they’re actually a second cousin once removed, or a first cousin twice removed. These relationship categories overlap in their DNA ranges, so some ambiguity is unavoidable.
At the bottom of your list, where matches share less than 15 cM, you’re increasingly looking at statistical noise rather than meaningful genealogical connections. That doesn’t mean every distant match is false. It means you need additional evidence before investing time in them. Shared matches (people who appear on both your list and a known relative’s list) can help confirm whether a distant connection is real. A family tree that intersects with yours in a specific location and time period adds further confidence.
The amount of shared DNA is also a better indicator of reliability than the relationship label itself. If you’re trying to verify a connection, focus on the centimorgan number rather than whether AncestryDNA calls someone your “4th-6th cousin.” A match sharing 45 cM is far more trustworthy than one sharing 8 cM, regardless of what label either one carries.