Google's AI Overviews YMYL results, the AI summaries shown on health and money searches, mostly cite pages that already rank well. In a 2026 study of 14,334 US health keywords, the No. 1 organic result was cited in 68.8% of AI Overviews. Ranking still matters most, but wrong answers on brand and support queries remain a separate risk.
This explainer is part of the generative engine optimization hub. YMYL (Your Money or Your Life) is Google's label for topics that can affect a person's health, financial stability or safety. An AI Overview is the generated summary Google shows above the classic results, with links to its sources, and a citation is one of those source links. The main recommendation: keep doing the ranking work, because it drives most citations in the new data, and add a monitoring routine for the queries where a wrong AI answer can hurt a customer.
Does Google treat YMYL queries differently in AI Overviews?
Google treats YMYL queries differently through its ranking systems, not through a separate rule for AI Overviews. Its helpful content guidance says its systems give even more weight to content with strong E-E-A-T on health, financial and safety topics, while its AI features page lists no additional requirements for AI Overviews.
E-E-A-T is Google's shorthand for experience, expertise, authoritativeness and trust. The Search Central guidance states that "trust is most important" and that E-E-A-T "itself isn't a specific ranking factor." For YMYL topics, it says content "must be highly accurate and consistent with established expert consensus." Google's overview of the rater guidelines defines YMYL as topics that "could significantly impact the health, financial stability, or safety of people." Rater data "is not used directly" in ranking, per the same Search Central page.
The AI features page says there are "no additional requirements to appear in AI Overviews or AI Mode": a page must be indexed and eligible to show with a snippet (as of November 2026). Health queries did get extra handling. In a May 30, 2024 post, Elizabeth Reid, VP of Search at Google, wrote: "In the case of health, we launched additional triggering refinements to enhance our quality protections." The SEO side of these rules is covered in the Google AI content guidelines explainer.
Who gets cited in AI Overviews YMYL results for health searches?
In AI Overviews YMYL results for US health searches, the domains that already rank win most citations. The performance.io and Semrush study found the No. 1 organic domain cited in 68.8% of snapshots with an AI Overview, and the organic top 10 holding 83.8% of all citations (PPC Land, 2026).
The study, announced on November 16 and 17, 2026, covered 14,334 unique keywords in five therapy areas from August 2022 to July 2026 (PPC Land, 2026). performance.io, a search agency, ran it on Semrush data. The citation rate falls steadily with rank: about 69% at position 1, 44% at position 3, 26% at position 5 and 13% at position 10 (PPC Land, 2026).
Community sites barely show up: Reddit received 0.15% of citations and ranked 80th, while YouTube received 1.76% and ranked 12th (PPC Land, 2026).
| Domain type | Citation share minus No. 1 share | What it suggests |
|---|---|---|
| Pharma brand sites | Minus 5.1 points (9.2% of No. 1 positions, 4.2% of citations) | Cited far below rank share |
| Government (.gov) | Minus 2.3 points | Cited below rank share |
| Disease associations | Minus 1.7 points | Cited slightly less |
| Major health publishers | Minus 1.0 points | Close to rank share |
| PubMed | Plus 0.9 points | Cited slightly more |
| YouTube | Plus 1.3 points | Cited above rank share |
| University (.edu) | Plus 1.7 points | Cited above rank share |
| Other domains | Plus 6.2 points | Smaller sites gain the most |
Why are pharma brand sites cited less often than they rank?
Nobody has shown why pharma brand sites are cited less often than they rank. The study's authors write that they "currently don't have an explanation" for the gap between 9.2% of No. 1 positions and 4.2% of citations (PPC Land, 2026). On their own brand queries, though, those sites do well.
When a pharma brand ranked No. 1 for its own brand query, the AI Overview cited the brand's domain 78% of the time, or 83.4% including sister sites, against 68.8% across all keywords (PPC Land, 2026). AI Overviews also spread on those queries: they appeared on 35 of 414 brand No. 1 snapshots in September 2024 and on 651 of 729 in July 2026 (PPC Land, 2026).
So the gap sits in non-brand searches about conditions and treatments. Possible explanations, none of them tested in the study, are labeled here as speculation: Google may prefer independent sources for general health answers, brand pages may be written for regulatory review rather than for patient questions, or brand pages may rank for terms they do not answer directly. Treat these as hypotheses to check on your own queries, not findings.
How much weight can the performance.io and Semrush study carry?
The performance.io and Semrush study carries weight as a large descriptive dataset, not as proof that rankings cause citations. It is observational, covers US desktop searches only, matches citations at domain level, uses modelled traffic, and was run by two companies that sell search services, as PPC Land noted in November 2026.
The press release itself says the study finds an association and does not establish that rankings cause citations (Business Wire via MarTech Edge, 2026). The authors call the work "exploratory and preliminary" (PPC Land, 2026). Five limits matter most:
- Sample size is smaller than announced. The 2026 release refers to more than 689,100 keyword snapshots. PPC Land found that 689,184 is the number of keyword-month slots requested, and only 340,660 returned data (49%, 2026).
- Domain, not page. A citation counts when the same domain appears, even if the cited URL differs from the ranking URL.
- No clicks. "Traffic figures are Semrush's modelled estimates, not measured clicks," in the authors' words.
- Desktop and US only. Mobile and other markets can differ.
- Commercial interest. Both companies sell related services, and Adobe agreed to buy Semrush in November 2025 (PPC Land, 2026).
The findings fit Google's rule that AI Overviews use pages eligible for regular results, but they do not show that a ranking gain produces a citation.
What goes wrong when AI Overviews answer health and support queries?
AI Overviews can show wrong health advice and fake customer service numbers, and both have been documented by the press. In August 2025 Android Central reported scam support numbers in AI Overviews, and in January 2026 Search Engine Journal covered a Guardian investigation into misleading health summaries.
In the phone number cases, a user searching for a Royal Caribbean 24-hour service line reached a scammer who obtained his credit card details, and a user searching for Swiggy customer care got a fake number, although Swiggy Instamart offers chat support only (Android Central, 2025). A Google spokesperson told The Washington Post the company was "aware" of the issue and working to "improve results," per the same report.
In the Guardian cases, Pancreatic Cancer UK called one dietary summary "completely incorrect," and Google answered that the "vast majority" of AI Overviews are "factual and helpful" (Search Engine Journal, 2026). Google's own AI Overviews help page says "AI Overviews can and will make mistakes." These are older, dated reports, not a measured error rate.
Which YMYL queries should a brand check first in AI Overviews?
A brand should check first the YMYL queries where a wrong AI Overview can cost a customer money or health: brand plus phone, support, cancel or refund, then brand plus dosage, fees or safety. The YMYL Exposure Grid below ranks query types by the harm a wrong answer causes, so monitoring starts where the risk is highest.
The YMYL Exposure Grid is a decision table, not a measured benchmark. It combines the reported failure modes with Google's YMYL definition of harm to health, finances or safety.
| Query type | Example pattern | Harm if the answer is wrong | Priority | First action |
|---|---|---|---|---|
| Brand plus contact | brand customer service number | Fraud, stolen payment details | 1 | Match every shown number with your official contact page |
| Brand plus money action | brand cancel subscription, brand refund | Lost money, wrong process | 1 | Publish the exact steps on one clear page |
| Brand plus safety | brand side effects, brand dosage | Health risk | 1 | Check against the approved label or terms |
| Non-brand condition or product | best savings account, type 2 diabetes treatment | Wrong decision, lost trust | 2 | Track who is cited and improve the ranking page |
| Non-brand comparison | brand A vs brand B fees | Lost sale, wrong facts about you | 3 | Correct facts on your own comparison page |
Run the grid on mobile and desktop.
How do health and finance brands protect themselves in AI Overviews?
Health and finance brands protect themselves in AI Overviews by keeping their ranking pages strong, publishing one official source for contact details, checking high-risk queries on a schedule and reporting wrong answers to Google. No step guarantees a citation, but each one shortens the time a wrong answer stays live.
- Keep the ranking work going. The organic top 10 took 83.8% of citations in the 2026 pharma data. See how to optimize for AI Overviews.
- Publish official contact details in two places. Put numbers on one crawlable contact page and in your Business Profile under "Edit profile", "Contact", "Phone number" (Business Profile Help); see local SEO.
- Run the YMYL Exposure Grid every month. Search the priority 1 queries, screenshot each AI Overview and log the cited domains, or use an AI Overviews tracker.
- Report wrong answers. Select the thumbs down icon under the AI Overview, choose "Report a problem", describe the error and select "Submit" (Google Search Help).
- Show who wrote and reviewed YMYL pages. Google asks whether pages "carry a byline, where one might be expected," and warns against "false credentials" (Search Central). Name real authors and reviewers.
- Measure citations next to clicks. The Generative AI performance report shows AI feature impressions per page; compare them with clicks using the AI Overviews click through rate method.
Common mistakes with AI Overviews on YMYL queries
The common mistakes with AI Overviews on YMYL queries come from reading one study as a rule and from ignoring the queries where a wrong answer causes real harm. Each mistake either wastes budget on the wrong fix or leaves a wrong answer about the brand live for weeks.
- Reading the pharma study as causal. The 2026 data shows an association. Test ranking changes on your own queries before you promise citations.
- Dropping SEO for a separate AI tactic. Google lists no extra requirements for AI Overviews. Ranking pages remain the main input.
- Checking only head terms. The reported scams sat on brand plus support queries. Add them to every monitoring list.
- Spreading contact numbers across old pages. Outdated numbers on press releases or PDFs give AI answers conflicting facts. Keep one official contact page and remove the rest.
- Inventing expert credentials. Google's helpful content guidance calls fabricated creator profiles a form of deception. Use real reviewers or leave the line out.
- Assuming Reddit drives health citations. Reddit took 0.15% of citations in the 2026 pharma data. Spend effort on your own pages and trusted health sources first.
