23andMe vs. AncestryDNA: Which DNA Test Wins for Health, Ancestry, and Privacy

23andMe vs. AncestryDNA is not a single contest: the right pick depends on whether you want health-risk reports or deep genealogical records. 23andMe sells FDA-authorized genetic health reports, including carrier status and polygenic risk scores, starting around USD 99 for ancestry alone and USD 199 with health. AncestryDNA targets family-history researchers, with a 26-million-person database and access to 100-plus million linked historical records, at USD 99 list (often discounted to USD 39-59). For health data, choose 23andMe. For finding ancestors and living cousins, choose AncestryDNA.

How each service actually works under the hood

Both companies collect a saliva sample, extract DNA in a lab, and read it on a genotyping microarray. The chips differ, and that difference explains most downstream divergence. 23andMe's current GSA chip genotypes roughly 640,000 markers, weighted toward variants with clinical relevance in the genes and regions its health models cover. AncestryDNA's chip reads about 700,000 markers chosen for population-genetics resolution, meaning variants that distinguish regional populations and detect distant kin.

Neither test sequences your genome. Both sample single-letter variants at known positions (SNPs) and infer the rest statistically. This is why neither can diagnose disease and why "raw data" downloads from the two are not interchangeable. Imputation—filling in untyped positions from reference panels—recovers common variants well but degrades for rare ones. A 2024 analysis of direct-to-consumer files found that re-imputing a 23andMe raw file to a standard reference panel recovers only about 80-90 percent concordance with typed data at clinically relevant loci, so cross-uploading to third-party tools (Promethease, GEDmatch) yields workable but imperfect results.

AncestryDNA's scale advantage is demographic, not biochemical: a 2024 figure puts it above 26 million tested customers versus roughly 15 million for 23andMe. That database is the product. Shared matches and predicted relationships depend directly on how many relatives have tested, and AncestryDNA has roughly 70 percent more people to match against.

In plain terms: both read a sliver of your genome, but one reads for health and the other reads for family connections.

The comparison table: health, ancestry, and price

The table below reconciles the headline differences. Note the pricing caveat: list prices are USD 99 for ancestry-only at both companies, but promotional pricing reshuffles the value calculus by season.

The single most important row is price-versus-database size. At USD 39 on sale, AncestryDNA delivers 70 percent more matches than a USD 99 base 23andMe kit. Conversely, no promotional discount closes the health-report gap, because 23andMe is the only one of the two with FDA-authorized health content.

What the health reports can and cannot tell you

23andMe's health tier splits into three distinct products, and conflating them is the most common buyer error. First, carrier status: tests for recessive conditions such as cystic fibrosis (CFTR variants) and sickle cell (HBB). These are highly predictive and clinically actionable. Second, polygenic risk scores (PRS) for conditions including type 2 diabetes, atrial fibrillation, and breast cancer. These are probabilistic, calibrated to specific populations, and less accurate for non-European ancestry. Third, pharmacogenomics: variants in genes like CYP2C19 and DPYD that affect drug metabolism.

A quantified caveat the marketing rarely stresses: a PRS in the top decile for a condition typically corresponds to a relative risk around 1.5-3 times the population average, not a diagnosis. A 2025 review of PRS performance found that prediction accuracy (measured as R²) for type 2 diabetes ranged from about 0.10 in European-ancestry cohorts down to under 0.04 in African-ancestry cohorts—roughly a 60 percent drop in explanatory power. Safety-critical numbers like these should be read against a clinician's judgment, not in isolation.

AncestryDNA does not report health at all through its own kit. Its 2022 acquisition of This is Me and later wellness partnerships push health into separate products and third-party integrations. If health is your goal, that gap alone settles 23andMe vs. AncestryDNA.

The non-obvious part: "ethnicity" means different math

Both companies advertise ancestry percentages, but the algorithms are not equivalent. 23andMe's ancestry composition uses a reference panel and reports confidence intervals—click any percentage and you see a range (for example, 12.4 percent French/German with a 5 percent lower bound). AncestryDNA reports point estimates with no visible confidence interval, though its "Genetic Communities" layer can place you in a hyper-specific group, sometimes pinning a region to a migration event in the 1700s.

The counterintuitive detail: 23andMe can assign you a maternal and paternal haplogroup because it types mitochondrial DNA and Y-chromosome markers directly on its chip. AncestryDNA does not report haplogroups at all. For deep paternal or maternal lineage tracing, 23andMe gives you a starting thread, but serious haplogroup research still requires a dedicated test (Family Tree DNA) because neither chip types enough Y or mitochondrial markers for fine resolution.

A second correction to a common misconception: larger database does not mean more accurate ethnicity percentages. AncestryDNA's 26 million customers improve cousin matching, but its ethnicity estimate draws on a reference panel of a few thousand individuals per region. 23andMe's regional breakdown, with 1,500-plus regions, can be finer in some areas of southwestern Europe and the Middle East, while AncestryDNA's African and Indigenous American coverage is generally regarded as stronger. The two disagree for 30-40 percent of regions on any given person, and both can be internally "right" against different reference panels.

Timeline: what happens after you spit in the tube

Processing differs by service and by season. The table below reflects typical 2025-2026 turnaround from kit registration to results, though holiday volume can stretch both.

If you are gifting a kit in late November, budget six weeks for results to arrive before a January deadline. The 2-3 week lab processing is the same order of magnitude at both; Ancestry's extra week is largely record-linking and community assignment.

Practical actions before you order or upload

The most consequential decision in 23andMe vs. AncestryDNA is not the test—it is what happens to your sample and data afterward. Both companies allow you to delete data, but deletion is not always complete or reversible, and third-party uploads create copies you cannot recall. Take these steps in order:

• Download raw data before making any account changes; it is your only portable copy.
• Read the research-consent screen line by line, especially auto opt-ins that share de-identified data.
• Use a privacy email alias and a unique password for the account.
• If you test with AncestryDNA and want health, do not simply upload to a third-party parser; have results confirmed at a clinical lab.
• Set a calendar reminder to re-check account and consent settings annually.
• Tell biological relatives before you test, since shared matches reveal family relationships they may not know about.

For buyers on the fence: the cheapest path to both goals is not one test but two—AncestryDNA for genealogy records and a discounted 23andMe health tier for carrier and PRS data. That costs roughly USD 140-240 in total and is the only combination that gets both the 26-million-person match pool and FDA-authorized health reports.

Limits, edge cases, and what neither test can do

Neither 23andMe nor AncestryDNA can detect most rare disease variants, structural rearrangements, or copy-number changes. Their chips miss small insertions and deletions and most variants outside the roughly 640,000-700,000 typed positions. A negative carrier report is not a medical clearance; carrier panels at a clinical lab cover more variants per gene and are ordered with genetic counseling.

Ancestry-specific edges are also real. AncestryDNA can find an adopted parent's lineage through cousin matching in cases where documents are missing, a scenario 23andMe handles less robustly because its smaller database produces fewer matches in the fifth-to-eighth-cousin range. Conversely, the 2025 bankruptcy sale of 23andMe raised a data-ownership edge case: purchaser control of a DNA dataset remains legally unsettled, and customers in some jurisdictions were given deletion windows before transfer. Treat any single-vendor loyalty as a risk, not a commitment.

Numbers should reconcile plainly: both chips type under 1 million of the roughly 3 billion base pairs—well below 0.1 percent of the genome—so a "complete genetic profile" claim from either company is marketing, not science. The practical rule is simple. Choose by your primary question: family tree, pick AncestryDNA. Health reports, pick 23andMe.