A guide to YMYL: where the concept comes from in Google’s guidelines, how it relates to E-E-A-T, what it changes in SEO, GEO and AEO, and a 24-item audit checklist.
By , founder and lead strategist at Flowup
Direct answer
YMYL (Your Money or Your Life) is the classification Google uses, in its Search Quality Evaluator Guidelines, for topics and pages that could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society: medicine, finance, law, safety, high-impact news and, since September 11, 2025, explicitly government and civic information as well (elections and voting). For these topics, the standard required is the strictest: the highest level of E-E-A-T, with verifiable authorship, primary sources, expert review and editorial transparency. YMYL is not a standalone ranking factor, but it defines how rigorously systems and human raters judge the page, and that same rigor is what decides whom AIs cite on sensitive topics.
YMYL stands for Your Money or Your Life: the classification Google uses for topics and pages whose content, if it is wrong, incomplete or ill-intentioned, can cause real harm to the health, financial stability, safety or well-being of people and of society. The definition is not market jargon: it is set out in the Search Quality Evaluator Guidelines (also known as the Search Quality Rater Guidelines), the official manual Google gives its quality raters, first published in full in 2015 and updated periodically since then. The current version, dated September 11, 2025, has 182 pages and is public.
The evolution of the concept tells the story of Google’s priorities. In August 2018, a core update that the trade press nicknamed the Medic Update hit health and wellness sites disproportionately, and the market understood that high-impact topics were judged by a yardstick of their own. In December 2022, the evaluation framework gained its first E, for Experience, valuing the real-life experience of the person who writes.
January 2025 brought the first revision since March 2024. And in September 2025, two changes signaled the current moment: the category previously called “YMYL Society” became “YMYL Government, Civics & Society”, explicitly including election and voting information, and the document gained examples of how to rate AI Overviews, which seem to mirror the examples already given for featured snippets. Google described the update as minor, with no change to its rating guidance, but the direction is unmistakable: AI answers and civic information have moved to the center of quality control.
The evaluation happens in two complementary layers, and understanding the split avoids two opposite mistakes: believing that “raters take sites down” and believing that the guidelines do not matter.
The first layer is the automated systems, which rank billions of pages by combining signals of relevance, quality, authority and consistency. The second is the human quality raters: around 16,000 people hired externally who follow the guidelines to rate samples of results.
The point the official documentation makes clear: the ratings given by raters do not change the ranking of any specific page; they work as a benchmark to measure whether the systems are promoting what the guidelines describe as quality, and to train the next versions of those systems. In practice, the guidelines are the document in which Google states what its algorithms try to reward.
For YMYL, that statement is explicit on three points. First, the standard required is the highest: YMYL pages need the highest level of E-E-A-T to receive high ratings. Second, the yardstick of consequence: the greater the potential harm of wrong information, the greater the rigor, and inherently dangerous topics (such as encouraging harm) receive the Lowest rating regardless of the source. Third, the 2025 update directs that content generated purely by AI, with no human review and no originality or added value, be rated as the lowest quality (Lowest), a direct message for the age of automated text factories.
The difference between YMYL and lower-risk content is not one of topic; it is one of consequence and, therefore, of standard. The table below, a Flowup synthesis, summarizes the eight dimensions in which the operation changes.
| Dimension | Ordinary content | YMYL content |
|---|---|---|
| Consequence of an error | Loss of credibility | Real harm: health, money, safety, rights |
| E-E-A-T standard required | Appropriate to the purpose of the page | The highest on the scale, with Trust at the center |
| Authorship | Desirable | Verifiable and qualified: name, credential, author page, professional registration where applicable |
| Sources | Recommended | Primary and cited in the paragraph itself: guidelines, regulations, studies, official documents |
| Review | Editorial | Expert and declared, with name, credential and date |
| Updating | When convenient | Scheduled and visible; immediate when the regulation or the guideline changes |
| Factual accuracy | A good approximation is accepted | Expert consensus is mandatory; an unsourced claim is a liability |
| Regulatory layer | Rare | Frequent. In Brazil: the CFM (Federal Council of Medicine) in healthcare, the CVM (securities commission) and Bacen (central bank) in finance, the OAB (Bar Association) in law |
YMYL classifies topics, not domains: the same site has fully YMYL pages and ordinary pages. The table below maps the categories of the guidelines with examples of pages and of real queries, taken from Brazil, Flowup’s home market, and translated into English.
| Category | Examples of pages | Examples of queries |
|---|---|---|
| Health and safety | Symptoms, treatments, surgeries, medications, mental health, vaccination, emergencies | “risks of surgery to stop wearing glasses”, “maximum dose of dipyrone” (a painkiller sold in Brazil), “heart attack symptoms” |
| Finance | Investments, taxes, credit, retirement, insurance, cryptocurrencies | “how to file an income tax return”, “is it worth taking an advance on the FGTS” (Brazil’s workers’ severance fund) |
| Legal | Rights, contracts, divorce, immigration, inheritance, consumer protection | “how does child support work”, “can I be fired while on medical leave” |
| News and high-impact events | Coverage of disasters, crises, public policy | “new unemployment insurance rule” |
| Government, civics and society | Elections, voting, public services, trust in institutions (made explicit in September 2025) | “how to justify not voting” (voting is mandatory in Brazil), “documents needed to vote” |
| High-impact purchases | Significant financial decisions, product safety | “mortgage or purchasing pool to buy a home” (the purchasing pool is a Brazilian group financing plan) |
| Groups of people | Information about ethnic, religious, gender and other groups | Informational queries about these groups |
The relationship between the two concepts fits in one sentence: E-E-A-T is the measure; YMYL defines how much of it is required. The Experience, Expertise, Authoritativeness and Trust framework is the vocabulary raters use to judge any page, with trust at the center: without Trust, the other signals do not sustain the rating. On ordinary topics, an adequate level is enough; on YMYL, what is required is the highest level, and first-hand experience (the first E, added in 2022) carries special weight: those who have lived it, treated it, operated or executed have a demonstrable advantage over those who have merely compiled.
A technical precision that sets this guide apart from common belief: Google’s official documentation states that E-E-A-T itself is not a ranking factor; the systems use mixes of measurable signals that approximate these concepts. The distinction matters so as not to fall for promises to “optimize E-E-A-T” as if it were a switch.
What does exist, and can be audited, are the signals: verifiable authorship, external reputation, factual accuracy, entity consistency, editorial transparency. They are what the work is done on, and they are the subject of the checklist at the end. Building these signals systematically, on and off the site, is the heart of Digital Authority Engineering: treating trust as infrastructure that can be designed, not as an accidental consequence.
In classic SEO, the effect of YMYL has been known since the Medic Update: core updates rigorously reassess the sensitive segments, and pages without strong trust signals lose visibility again and again. What changed in 2025 and 2026 is that the same standard migrated to the answer layer.
Three movements define the landscape. First, AI Overviews became a formal object of evaluation: the September 2025 guidelines bring examples of how to rate AI summaries, which seem to mirror the examples given for featured snippets, and the presence of these summaries reached 48% of the queries tracked by BrightEdge in February 2026. Second, the click shrank where AI answers (8% against 15%, in the Pew Research panel with 68,879 real searches), which makes the citation inside the answer the new asset in dispute, as we detail in the comparison GEO vs. SEO. Third, the selection of sources by generative systems favors exactly the attributes of the YMYL standard: the Princeton study (KDD 2024) measured visibility gains of up to 40% for content with statistics, cited sources and clarity, and the full mechanics of retrieval, selection and attribution are in our complete GEO guide.
The strategic synthesis: on YMYL topics, GEO and AEO are not disciplines separate from trust; they are its reward. Answer engines, with more to lose when they get health or money wrong, anchor themselves in sources with verifiable authorship, primary data and entity consistency. In Brazil, there is also the regulatory layer we analyze in the guide to YMYL and medical advertising under the CFM (in Portuguese): CFM Resolution 2,336/2023, from Brazil’s Federal Council of Medicine, requires identification (name, CRM, the medical council registration number, and RQE, the specialist registration number) and prohibits promises of results, requirements that converge with what the YMYL standard rewards. Whoever builds for the regulation builds, at no additional cost, for the algorithm.
1. Anonymous or ghost authorship. Health or finance content “by the Editorial Team” is disqualifying under the current standard. 2. Unsourced claims. In YMYL, a sentence without a reference is not neutral: it is an auditable liability. 3. Promise of results. A guarantee of cure, of financial return or of victory in court violates the trust standard and, in Brazilian healthcare, the CFM rules. 4. Outdated content treated as current. A revoked regulation, a superseded guideline or an old rate presented as current.
5. Internal contradiction. Pages of the same site stating different things about the same fact, the opposite of entity consistency. 6. AI without review. Publishing generated text without expert validation, exactly the profile the guidelines tell raters to rate Lowest. 7. Decorative schema. Structured data declaring credentials, reviews or ratings that the visible page does not support. 8. Ads over content. Aggressive monetization burying the main content, a classic cause of lower quality ratings in the guidelines.
The trust architecture of a YMYL operation has four layers, and all of them need to be readable by people and by machines.
Authorship and review. Every piece of content with a real, credentialed byline; a complete author page (education, professional registration, experience, external profiles); and declared expert review with name, credential and date. The writer plus technical reviewer arrangement is the most defensible at scale: it separates the editorial function from the expert responsibility. Editorial transparency. A published editorial policy (how content is produced, reviewed and corrected), sources cited in the paragraph of the claim, publication and review dates that are visible and true, and a channel for corrections.
Structured data. A coherent graph with Organization, Person for authors and reviewers (with sameAs pointing to official profiles), specific types where applicable and FAQPage mirroring real questions, always with absolute coherence between markup and visible content, following the standard we detail in the guide to JSON-LD structured data. Content and entity architecture. The layer that ties everything together: an information base that centralizes the canonical facts of the brand and of its experts (the source of truth that an official knowledge base for AIs (in Portuguese) makes concrete), territories covered in depth with pillar pages and satellite pages, and content that adds to the corpus instead of repeating it, the principle of information gain that also protects against copying. In the B.I.N.A. Method, these layers are not separate projects: they are fronts of a single engineering effort, from the semantic information base to the consolidation of the entity, with external validation fed by Data-Driven PR.
Use the four blocks below as an audit: each item not met is a hypothesis for why the page does not perform, is not cited or does not turn trust into business.
The measurement process for the last layer, including the prompt panel and the Search Console filter for AI features, is detailed in the guide on how to measure visibility in AI answers.
YMYL stands for Your Money or Your Life. It is the classification Google uses, in the Search Quality Evaluator Guidelines, for topics and pages that could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society. Since September 2025, the definition explicitly includes government and civic information, such as elections and voting. For this content, the quality standard required is the highest on the scale.
The core categories are: health and safety (symptoms, diagnoses, treatments, medications, mental health, procedures, emergencies); finance (investments, taxes, credit, retirement, insurance); legal topics (rights, contracts, lawsuits, immigration); news and high-impact events; government, civics and society (elections, voting, public services, trust in institutions, a category made explicit in 2025); high-value purchases; and information about groups of people. The yardstick is consequence: if an error can cause real harm to someone’s life, the topic is YMYL.
Not directly. YMYL is a classification used by the human quality raters and reflected in Google’s systems, not a standalone factor that is switched on or off. Rater ratings do not change the rankings of specific pages; they serve to measure and train the systems. The practical effect, however, is real: on YMYL topics the systems require much stronger signals of experience, expertise, authoritativeness and trust, and pages without these signals perform consistently worse.
The concept was born in the Search Quality Evaluator Guidelines, the manual Google gives its external quality raters. The document was officially published in full in 2015 and is updated periodically; the current version, dated September 11, 2025, has 182 pages. In August 2018, a core update that the trade press nicknamed the Medic Update disproportionately hit health sites and other YMYL segments, and the SEO market came to treat the concept as central.
Two updates marked the year. In January 2025, Google revised the guidelines for the first time since March 2024. On September 11, 2025, the category previously called YMYL Society became YMYL Government, Civics & Society, explicitly including election and voting information, and the document gained examples of how to rate AI Overviews, which seem to mirror the examples given for featured snippets. Google described the change as minor, with no change to its rating guidance, but it signals the two priorities of the moment: AI answers and civic information.
E-E-A-T (Experience, Expertise, Authoritativeness, Trust) is the framework raters use to judge the quality of any page; YMYL defines how much of that standard is required. On ordinary topics, a reasonable level of E-E-A-T is enough; on YMYL topics, what is required is the highest level, with Trust at the center: without it, the other signals do not sustain the page. The operational consequence: in YMYL, verifiable authorship, credentials, primary sources, expert review and editorial transparency stop being differentiators and become requirements.
Not as a direct factor. Google’s official documentation states that E-E-A-T itself is not a specific ranking factor; the systems use a mix of signals that serve as proxies for experience, expertise, authoritativeness and trust. In practice, for those who produce content, the distinction is academic: the signals the systems can measure (authorship, reputation, citations, consistency, accuracy) are exactly the ones E-E-A-T describes, and in YMYL they weigh more.
Two layers. The first is Google’s automated systems, which combine signals of quality, authority and consistency at scale. The second is the human quality raters, around 16,000 people hired externally, who follow the Search Quality Evaluator Guidelines to rate samples of results. Their ratings do not change the ranking of individual pages: they work as a benchmark to measure and train the systems. On YMYL topics, both the systems and the raters apply the highest standard.
It can be used as a tool, not as the final author. The September 2025 guidelines instruct raters to rate as the lowest quality (Lowest) content generated purely by AI, with no human review and no originality or added value. In YMYL, the risk is multiplied: a factual error generated at scale can cause real harm. Responsible use combines AI for structuring and research with qualified human authorship, documented expert review, primary sources and proprietary data, exactly the standard that also produces information gain.
It affects them in two directions. First, AI Overviews are now formally rated: the September 2025 guidelines include examples of how to rate AI summaries, which seem to mirror the examples given for featured snippets, and the standard rises on sensitive topics. Second, generative systems select sources by signals of trust and verifiability, which favors exactly those who meet the YMYL standard: real authorship, sourced data, entity consistency. On health and finance topics, being citable requires being trustworthy first.
It is not the whole site that receives the classification; it is the topic of each page. At a clinic, the page about a surgical procedure, its risks and the postoperative period is fully YMYL; a company post about the team’s year-end party is not. At a bank, the page that explains how to invest is YMYL; the job openings page, in general, is not. The common mistake is to treat everything with the same shallow rigor, instead of concentrating the highest standard on the pages where the consequence of an error is real.
Medic Update is the nickname the trade press gave to the August 2018 core update, which disproportionately hit health and wellness sites and other YMYL segments. It matters as a historical milestone: it was the moment the market realized that Google treated high-impact content with a distinct standard, and that signals of expertise and trust weighed more on these topics. Since then, each core update tends to rigorously reassess the YMYL segments.
With verifiable identity in three layers: a byline with a real name and a relevant credential on the page itself; a complete author page, with education, professional registration where applicable, experience and links to external profiles; and Person structured data connecting the author to the content, with sameAs pointing to the official profiles. In healthcare, the Brazilian standard adds a regulatory requirement: name, CRM (the medical council registration number) and RQE (the specialist registration number) in publications, under CFM Resolution 2,336/2023, from Brazil’s Federal Council of Medicine. Ghost authorship in YMYL is an invitation to distrust from systems and readers.
It is the validation of the content by a professional qualified in the topic, separate from the writing. A correct declaration identifies who reviewed it, with credential and date (for example: clinically reviewed by, with the CRM and RQE registration numbers used by physicians in Brazil, on a given date), in a visible place on the page, and can be reinforced in the structured data. The writer plus expert reviewer pair is the most defensible arrangement for YMYL operations at scale: it separates the editorial function from the technical responsibility and creates an audit trail.
It helps in the layer of understanding and consistency, which is where trust begins for machines. Organization schema with verifiable data, Person for authors and reviewers, MedicalWebPage or specific types where applicable, FAQPage mirroring real questions and a graph with stable identifiers reduce ambiguity about who states what. What structured data does not do is compensate for weak content: schema declaring credentials that the visible content does not support is noise, and coherence between markup and page is a Google requirement.
The tolerance for error. In ordinary content, an inaccuracy costs credibility; in YMYL, it can cost someone’s health, money or rights, and so the standard required jumps: credentialed authorship instead of anonymous writing, primary sources instead of common belief, review dates instead of static content, factual accuracy instead of approximation, and the highest level of E-E-A-T instead of a reasonable level. The comparison table in this guide details the differences across eight operational dimensions.
They can, when they have real expertise and demonstrate it. The YMYL standard does not measure size; it measures reliability: a specialist with verifiable credentials, deep content in a niche, proprietary data and entity consistency can outperform generic portals in their own territory. The project of Dr. Ana Vega in Brazil, documented by Flowup, shows a solo healthcare practice building a dominant organic presence in its niche with this standard. What does not work in YMYL is being small and generic at the same time.
As often as the facts change, with scheduled review even when nothing seems to have changed. A practical recommendation: expert review every 6 to 12 months for stable clinical and financial content, immediate review when there is a regulatory change (a new resolution, a new law, a new clinical guideline) and quarterly review for market data. The last-review date must be visible and true: dates touched up without any real change are the kind of signal that undermines the trust you are trying to build.
The most destructive patterns: health or money claims without a source; anonymous or generic authorship; promises of results (a guarantee of cure, of financial return, of victory in court); outdated content treating a revoked regulation as current; contradiction between pages of the same site; excessive advertising over the main content; schema declaring what the page does not show; and AI-generated content published without review. Each of these signals, on its own, already lowers the rating; combined, they put the page in the lowest-quality group.
Yes, and with growing force. GEO and AEO compete for presence inside AI answers, and generative systems select sources by verifiability, authority and consistency, exactly the signals the YMYL standard requires. On health and finance topics, AIs tend to ground their answers in institutional and specialized sources, and the expansion of the guidelines to rate AI Overviews formalized that rigor. The strategic reading: the investment in YMYL trust is the same investment that makes the brand citable by answer engines.
With a diagnosis in four questions: which pages of the site are in fact YMYL; what they state without a source, without an author or without a date; how the entity (brand and experts) is recorded on and off the site; and what AIs answer today about the topic and about the brand. From that picture come the priorities, in the order of this guide’s checklist: authorship and trust, content and evidence, technical layer and maintenance routine. Flowup’s B.I.N.A. Diagnosis carries out this initial mapping, including the AI visibility layer.
Next step
If your operation touches health, money, law or decisions that affect lives, trust is not rhetoric: it is auditable infrastructure. The B.I.N.A. Diagnosis maps your YMYL pages, the E-E-A-T signals that are missing, the consistency of your entity and what AIs answer about your topic, and returns the priority plan. Ranking is not enough. Be the answer.
Founder and lead strategist, Flowup Agency
Guto Bertoncini is the founder and lead strategist of Flowup Agency, which he has run since 2011. He is the author of the B.I.N.A. Method, Novo SEO and the Base Informacional Semântica (Semantic Information Base), and leads the agency's SEO for AI, GEO and AEO practice, preparing companies to be found on Google and cited by artificial intelligence platforms. He writes about search and AI on the Flowup blog and on his official website.
Sources verified on August 5, 2026. The statements about the Search Quality Evaluator Guidelines reflect the version of September 11, 2025 (182 pages), according to the public document and specialized coverage; the guidelines guide human raters and do not directly change rankings, as Google’s documentation states. E-E-A-T is not a direct ranking factor, according to the same documentation. The market data cited (BrightEdge, Pew Research, Princeton KDD 2024) use different methodologies and time windows, stated in the text.
The indicators of Dr. Ana Vega’s project are public, with approximate channel attribution and no guarantee of repetition; Dr. Alexander Kopelman’s project is at an early stage and is cited without indicators, since there is no consolidated public data. Statements that require periodic review: the current version of the guidelines and its YMYL categories, the percentages of AI Overview presence, the status of CFM Resolution 2,336/2023 and the indicators of the projects cited. No data was invented; where the evidence was insufficient, the data point was left out.
