How AI Search Is Recommending DNA Testing Kits: Monthly Trends

Executive Summary

AncestryDNA remains the category leader in October 2026 with valid recommendation coverage of 73.2%, holding a lead of 5.9 points over second-place 23andMe at 67.3%. The leadership position is unchanged from the baseline, and the leader's coverage moved by less than a single point from the immediately prior month, keeping the top of the category stable. The most consequential shift in the current month is lower down: Nebula Genomics fell to 16.5% coverage, a significant decline of 10.8 points from July's 27.3%, the largest baseline decline in the category. FamilyTreeDNA also registers as a significant decliner, down 6.4 points to 57.7%.

The strongest upward movement in the current month came from brands that remain below the leadership tier. 23andMe rose 2.2 points from the prior month to 67.3%, though the move did not break the normal range of month-to-month variation. MyHeritage DNA, by contrast, fell 1.9 points from September's 61.4% to 59.5% in October, continuing a month-to-month decline even as its placement metrics improved. On the placement side, MyHeritage DNA's top-three rate is the standout riser across the series, up 5.8 points to 49.7%, a significant gain. No brand recorded a significant rise in coverage this month.

Against the full July-to-October series, the category's defining movement is downward at the mid and lower tiers. Four brands, FamilyTreeDNA, Health Nucleus, Nebula Genomics, and tellmeGen, are classified as significant decliners measured from baseline. The leader by coverage has not changed in any month of the series, and the month-to-month reading is best described as stable at the top with continued erosion beneath it.

Seven observations were classified as irrelevant in the qualification funnel this month, and the qualified benchmark set contracted to 563 observations, the smallest of the four months, down from 614 in September.

Each monthly run begins with 800 prompt-surface observations (399 unique questions in July, 455 in September, 467 in October) across the benchmark's defined AI/search surface universe. Of those, 800 mentioned a tracked brand or competitor in each month; 799 were relevant and 1 irrelevant in July, and 793 were relevant with 7 irrelevant in October. The public metrics use the 594 qualified observations in July and the 563 qualified observations in October that survive both qualification stages.

Key Findings

AI Response Inconsistency Alerts

Questions This Section Answers

• Which companies had the most factual inconsistencies across AI platforms?
• What unresolved conflict affects how AI describes 23andMe's corporate status?
• Where did platforms disagree about free raw DNA upload policies?

The benchmark detected 23 critical and high-severity factual inconsistencies across seven AI platforms. These are cases where different AI surfaces gave directly incompatible answers to the same or near-identical questions about the same company. The conflicts below are grouped by the company they concern.

23andMe

Ten inconsistencies were logged for 23andMe, the highest count of any company in the vertical, and most concern its corporate status. When asked "Which is the best DNA kit for ancestry?", Google AI Mode described 23andMe's situation as recent corporate restructurings with no mention of bankruptcy, while ChatGPT stated that 23andMe filed for Chapter 11 bankruptcy in 2025. Google AI Mode cited Wirecutter, CNET, and Genomelink, while ChatGPT referenced a Reuters report on the filing.

The divergence repeats across question themes. Asked "What is the most reliable DNA testing company?", Gemini presented 23andMe as the gold standard for health and wellness insights and recommended it as reliable, whereas Copilot stated that 23andMe suffered a major data breach and bankruptcy in 2025, raising trust issues, citing a Yahoo Lifestyle page that noted the company had been removed from a best-DNA-test list. Asked "Which DNA website is the most accurate?", Google AI Mode pointed to 23andMe's own ancestry page and recommended it without qualification, while Copilot again stated that 23andMe went bankrupt in 2025 and left 15 million users' data in limbo, citing a DNA test comparison table.

Two further conflicts centred on the cheapest alternative query and the best-company query. On "What is the cheapest alternative to 23andMe?", Gemini implied 23andMe was operating normally with no bankruptcy mention, while Copilot stated the company went bankrupt in 2025, drawing on a buyer's guide that referenced the March 2025 Chapter 11 filing. On "What is the best company to get a DNA test from?", ChatGPT stated that 23andMe went through bankruptcy after a 2023 data breach and that its genetic-data assets were sold to TTAM Research Institute, citing a Michigan Attorney General release, while Copilot presented 23andMe as still the most established option and the leader for health and ancestry. Similarly, on "What are the best DNA websites?", Copilot flagged that 23andMe went bankrupt in 2025 with an uncertain future, while ChatGPT presented it as a current active top option with no bankruptcy reference.

The pattern continues on home-test and accuracy queries. Asked "What is the best DNA test to take at home?", Gemini described 23andMe as an active recommended health and ancestry company, while Copilot stated it had a bankruptcy in 2025. Asked "Which company is the best for genetic testing?", ChatGPT presented 23andMe as a current recommended option with no instability caveat, while Copilot (high severity) stated it underwent bankruptcy restructuring in 2025 and that long-term support is less certain. On "What DNA test is more accurate?", Google AI Overviews presented 23andMe as the market leader for FDA-authorized health reports, while Copilot noted the bankruptcy and acquisition and flagged stability as a consideration.

One pricing conflict was logged for 23andMe. On the query "dna test how much", Google AI Overviews stated an Ancestry plus Health price range of $119 to $229, while Google AI Mode listed the same Health + Ancestry product at $100 to $640 under health and wellness assessments.

Nebula Genomics

Three critical inconsistencies were logged for Nebula Genomics, all concerning whether the company is still operating. On "What is the highest rated DNA test kit?", Gemini stated that Nebula Genomics is active and recommended for whole genome sequencing, citing Nebula's own blog, while Copilot stated the company shut down in 2025 and that Sequencing.com is the successor, citing a buyer's guide that named a replacement provider.

The conflict repeats on accuracy queries. Asked "What is the most accurate DNA test kit available?", Gemini described Nebula Genomics as a top kit for whole-genome health analysis and currently operating, while Copilot stated it shut down in 2025 and was replaced by DNA Complete. On the shorthand version of the same question, Gemini again described Nebula as an active provider of 30x whole genome sequencing with no source URLs attached, while Copilot stated it shut down in 2025.

MyHeritage DNA

Five inconsistencies were logged for MyHeritage DNA, all concerning whether the company still accepts free raw DNA uploads from other testing services for relative matching. On "Is MyHeritage DNA worth the money?", Google AI Mode stated that MyHeritage transitioned away from allowing new users to upload raw DNA data for free matching and that a kit purchase is now required, citing a KnowYourDNA review that noted raw DNA uploads from other services are no longer accepted. Copilot stated the opposite, that raw data can be uploaded for free to expand a match pool if a user has already tested with AncestryDNA or 23andMe.

The conflict recurs across several ancestry queries. On "Which is the best DNA kit for ancestry?", Google AI Mode stated that free raw DNA uploads are no longer accepted, while Gemini stated that a raw DNA file can be uploaded for free or for a small unlock fee. On "What is the best Ancestry test website?", Google AI Mode stated that new users must purchase a kit and that free raw data uploads for matching are no longer supported, while Gemini stated raw data can be uploaded for a low fee and Copilot stated that raw data uploads from other services are supported. A fourth instance on the same query had Google AI Mode restate that free uploads are discontinued while Google AI Overviews stated that a raw DNA file can be uploaded for free or a small fee to MyHeritage, FamilyTreeDNA, or GEDmatch.

AncestryDNA

Two high-severity pricing inconsistencies were logged for AncestryDNA. On the query "ancestry dna test kit", Google AI Overviews stated that kits start around $34 to $39 for basic ethnicity tests, citing a Facebook promotion, while Google AI Mode stated the kit is typically priced around $99 for the basic kit. On "What is the best DNA ancestry kit to buy?", Google AI Overviews stated a price of around $99, frequently on sale, while Copilot reported a current price of $212.49 alongside a separate $38.00 listing, citing third-party retail listings.

Living DNA

Two high-severity pricing inconsistencies were logged for Living DNA. Asked "How much does Living DNA cost?", Google AI Overviews stated the Full Ancestry Kit costs $99.00, while Google AI Mode stated the Ancestry Kit costs $124. On the same query, Google AI Overviews priced the Ancestry and Wellbeing Kit at $169.00, while Google AI Mode stated the Wellbeing and Ancestry Kit costs $179 to $225, with source pages on Living DNA's own site showing both $169 and $179.

FamilyTreeDNA

One high-severity inconsistency was logged for FamilyTreeDNA, concerning law enforcement matching. Asked "Is FamilyTreeDNA trustworthy?", Gemini stated that users can explicitly opt out of law enforcement matching and that opting out hides their profile, while Google AI Mode stated that law enforcement matching is strictly opt-in and that profiles are hidden unless a user explicitly participates. These describe opposite default states for the same program.

Benchmark Context

The report separates the raw collection universe from the qualified analysis set. Brand-level recommendation percentages are calculated within the qualified benchmark set.

August and September sat between these points: 800 total prompts and 607 qualified observations in August, and 800 total prompts and 614 qualified observations in September. The qualified benchmark observation count is the denominator for every brand-level metric below, so the smaller October set means each brand's coverage percentage is calculated across fewer observations than in any prior month of the series.

Benchmark-Level Metrics

The recommendation-shaped answer share moved unevenly before reaching October: it fell to 36.1% in August, recovered to 40.1% in September, then rose sharply to 49.4% in October, its highest level in the series. The valid recommendation shortlist share followed a gentler path, dipping to 69.5% in August and 70.8% in September before returning to 74.4% in October. The rise in recommendation-shaped answers alongside a smaller qualified set indicates that a larger proportion of this month's observations produced recommendation-style output.

AI Recommendation Trend

Questions This Section Answers

• Which brand leads the category across the July-to-October series?
• Which brands recorded significant declines from the July baseline?
• How much did AncestryDNA's coverage shift, and did its lead change?

A Steady Leader With Persistent Erosion Beneath the Top Tier

Against the baseline, four brands exceeded the normal range of month-to-month variation: FamilyTreeDNA, Health Nucleus, Nebula Genomics, and tellmeGen. Against the immediately prior month, no brand moved significantly, so the current-month category change reflects the combination of several smaller movements rather than a single break.

What Changed This Month

Questions This Section Answers

• What drove Nebula Genomics' largest decline in the category?
• How did MyHeritage DNA improve its top-three placement rate?
• Why did FamilyTreeDNA lose mention presence faster than recommendation placement?

AncestryDNA

AncestryDNA's valid recommendation coverage stands at 73.2% in October, down 3.7 points from July's 76.9% but effectively flat from September's 73.3%, a change of 0.1 points. Its leadership position is unchanged for the fourth consecutive month.

Placement measures show the same stability at the top of the funnel. The brand's rank-one rate sits at 69.6% in October against 69.9% in July, with 392 rank-one placements, and its top-three rate is 73.0% against July's 75.1%. AncestryDNA recorded 411 top-three placements against 392 rank-one placements in October, so the brand placed first in the large majority of its top-three appearances.

The distinction worth noting is that AncestryDNA's coverage is close to its raw mention presence, which sits at 99.6%. The brand is mentioned in almost every qualified response and recommended in roughly three of every four, so its standing rests on both near-universal presence and consistently high placement rather than on presence alone.

Highest-priority diagnostic: Which surfaces or prompt types account for the small residual gap between AncestryDNA's 99.6% presence and its 73.2% coverage?

23andMe

23andMe's valid recommendation coverage is 67.3% in October, down 3.7 points from July's 71.0%. The prior-month comparison is the more favorable read: coverage rose 2.2 points from September's 65.1%, the largest upward move of any brand this month.

The placement picture is mixed. The brand's top-three rate fell 5.9 points over the series to 61.6% from July's 67.5%, a significant decline, even as coverage held up. Its rank-one rate slipped 1.4 points to 1.6% with 9 rank-one placements, down from 18 in July. Top-three placements fell from 401 in July to 347 in October.

The key distinction is between coverage and placement. 23andMe remains present in 97.2% of October responses and is recommended more often than any brand except the leader, but it is being placed lower within recommendations. Its valid recommendation count of 379 is down from 422 in July.

Highest-priority diagnostic: Which prompt themes moved 23andMe from top-three placements to lower positions, and which brand took those slots?

FamilyTreeDNA

FamilyTreeDNA's valid recommendation coverage is 57.7% in October, down 6.4 points from July's 64.1%, a significant decline across the series and the fourth-ranked position in the category. The move from the prior month was gentler at 1.1 points.

The divergence here is instructive. Coverage fell, but the top-three rate actually rose 0.8 points over the series to 25.4%, and the rank-one rate fell 0.8 points to 0.2% with a single rank-one placement. Raw mention presence declined 4.3 points to 75.7% from July's 80.0%, while top-three placements fell only slightly, from 146 in July to 143 in October, despite the smaller qualified set.

The distinction is that FamilyTreeDNA is losing mention presence faster than it is losing recommendation placement. It appears in fewer responses, but when it does appear, its top-three rate has edged up rather than down, suggesting the loss is concentrated in responses where it was mentioned but not recommended.

Highest-priority diagnostic: Which question types stopped mentioning FamilyTreeDNA at all, and do those same prompts now mention a competitor instead?

MyHeritage DNA

MyHeritage DNA's valid recommendation coverage is 59.5% in October, down 2.3 points from July's 61.8%, a move within normal variation. Coverage fell 1.9 points from September's 61.4%, continuing to sit just above FamilyTreeDNA in the rankings.

Placement is the stronger story for this brand. Its top-three rate rose 5.8 points over the series to 49.7% from July's 43.9%, a significant gain and the largest positive placement movement in the category. Top-three placements rose from 261 in July to 280 in October. The rank-one rate fell 0.5 points to 0.9% with 5 rank-one placements.

The distinction is that MyHeritage DNA is being recommended in fewer responses overall but is placing higher when it is recommended, with a top-three rate now approaching half of all qualified observations.

Highest-priority diagnostic: Which prompts drove the top-three improvement, and does it hold across surfaces?

Nebula Genomics

Nebula Genomics' valid recommendation coverage is 16.5% in October, down 10.8 points from July's 27.3%, the largest significant decline in the category and the sharpest baseline movement of any brand. It has declined in every month of the series: 27.3% in July, 21.8% in August, 20.0% in September, and 16.5% in October.

The decline is broad-based. Raw mention presence fell 7.9 points to 22.4% from July's 30.3%, a significant drop. The top-three rate fell 2.7 points to 7.1%, with top-three placements down from 58 to 40. The valid recommendation count fell from 162 in July to 93 in October.

The distinction is that Nebula Genomics' rank-one rate actually rose 0.4 points over the series to 1.2%, even as coverage, presence, and top-three rates all fell. With only 93 valid recommendations, this brand's percentages can move materially on a change of a few observations, so the rank-one gain should be read alongside the counts rather than in isolation.

Highest-priority diagnostic: Which surfaces and question types account for the reduced mention presence, and is the rank-one retention concentrated in a small number of prompts?

tellmeGen

tellmeGen's valid recommendation coverage is 0.5% in October, down 2.0 points from July's 2.5%, a significant decline measured from baseline with only 3 valid recommendations in the month. Raw mention presence fell 2.3 points to 2.1% from July's 4.4%, a significant drop. The brand retained 1 rank-one placement and 2 top-three placements.

The distinction is one of scale. With an observation count this small, a shift of a single prompt changes the percentage materially, and the decline should be read as directional rather than as a precise rate. The brand's net sentiment score also fell to 0.2 from July's 0.6.

Highest-priority diagnostic: Which specific prompts still surface tellmeGen at all, and do those prompts overlap with any competitor's losing prompts?

Buyer-Intent Interpretation

Questions This Section Answers

• Which buyer-intent clusters produced qualified observations in the benchmark?
• What can the benchmark currently say about how AI positions brands on pricing or value?

Across July through October 2026, all qualified observations fell into the Brand Recommendation cluster. The public benchmark captured no qualified observations in the Pricing & Value or Multi-Brand Comparison clusters in any month of the series, even though pricing and comparison response types appear in the raw distribution. The public metrics therefore answer the "which brand" question well, but they cannot yet speak to how AI positions these brands on price, value, or direct head-to-head comparison. That limitation is notable given that the inconsistency data surfaced several pricing disagreements between surfaces, which sit outside the qualified benchmark's measurable scope.

Brand Opportunity Summary

Questions This Section Answers

• Which brands have the highest-priority diagnostic for their current visibility signal?
• What is the key opportunity signal for each brand in October?

The benchmark identifies where attention is warranted; a company-level analysis is needed to explain why.

Evidence Behind the Benchmark

The aggregate metrics are built from prompt-level observations (query, surface, recommendation outcome, rank, sentiment, and citations where exposed). Company-level analysis can go deeper into prompt, competitor, surface, and evidence patterns. Source presence is not automatically treated as proof of causation.

About This Benchmark

This report is part of the CiteWorks Studio AI Visibility Market Discovery research program.

• AI Visibility Industry Market Methodology
• AI Visibility Industry Market Metrics
• AI Visibility Industry Market Standards

Report-Specific Interpretation Notes

• Small-count movement: brands with fewer than 10 valid recommendations (tellmeGen with 3, CRI Genetics with 3, and Health Nucleus with 0 in October) can move meaningfully with a change of just a few prompts; treat their percentage shifts as directional rather than precise.
• Qualified denominator: all brand-level percentages are calculated within the qualified set of 563 observations in October, not the 800 raw prompts collected.
• Directional analysis: month-over-month movement identifies changes worth investigating; it does not by itself establish the cause of those changes.

Next Step

The Public Benchmark Shows Where a Brand Is Winning or Losing. A Company-Level Audit Shows Why.

Beneath the aggregate percentages sit questions the public benchmark cannot answer: which high-intent prompts a brand wins, which competitor takes the recommendation when a brand loses, what attributes AI associates with each option, and which external sources shape those answers. The significant baseline declines at FamilyTreeDNA, Nebula Genomics, and tellmeGen, alongside MyHeritage DNA's placement gain, raise specific questions about which prompts and surfaces drove those shifts across the four-month series.

A company-specific AI visibility audit maps those prompt, surface, competitor, ranking, sentiment, and evidence-source patterns into a prioritized visibility strategy. It turns the benchmark's "what changed" into an actionable "why and what next."

Request an AI visibility audit