YMYL

What it is

Some pages can hurt someone. That is the whole basis of the category, and it is a more sensible basis than most things in search documentation.

Google gives its human quality raters a manual, the Search Quality Rater Guidelines, and asks them to judge sample results. The manual distinguishes topics on which a bad page is merely unhelpful from topics on which a bad page can do damage to a person's body, money, safety or future, and it instructs raters to apply the strictest scrutiny to the second kind. Medical advice, financial advice, legal information, safety information, and pages that bear on a major life decision all sit inside it. A page about which typeface to use does not.

The dates, verified in the vault's pass of 1 September 2026: the concept sits in Google's rater documentation from 2013, was published openly in 2014, and its definitions were clarified again in September 2025. That last date is worth carrying, because a great deal of published guidance on the topic predates it and describes a scheme that has since been re-scoped.

Two things this is not, and both errors are common enough to be worth stating before anything else.

It is not a ranking factor. Nothing in the ranking system reads a YMYL flag, no such flag exists to be read, and the raters do not change the position of the pages they score. Their judgments are used to test whether algorithm changes are working, which makes them an instrument rather than a control. Anyone selling YMYL optimisation is selling a thing without a mechanism.

It is not a research finding either, which is what the not-a-research-claim status records. It is a published classification scheme belonging to a company, checkable against that company's own documentation and against nothing else. Replication does not apply. It can be described accurately or inaccurately, and it can change when Google decides to change it.

What it actually does is impose asymmetric editorial cost. A publication writing about attention and consumer behaviour can be wrong about the sharing habits of New York Times readers and cost a reader nothing. The same publication, wrong about attachment, about training loads, or about eating, can cost a reader something that is not recoverable by publishing a correction. The classification is Google's, but the reason behind it is not really about search at all: it is a restatement of the ordinary editorial principle that the standard of care should scale with the harm.

That principle is why this site treats YMYL as a production constraint rather than as an SEO consideration. The requirements it triggers here are specific: citations to primary sources rather than to aggregators, credentialed reviewers on clinical material, visible disclaimer blocks, and a standing decision to stay out of the heaviest territory, suicide, medication and severe mental illness, until professional review is in place. None of those is a schema field. All of them cost time, which is the point.

There is a sharper reason for a psychology publication to care. The material most likely to be searched for by someone in difficulty is also the material where popular psychology has the worst citation habits, and where a confidently written page repeating a superseded finding does the most harm. The overlap between the YMYL category and the replication crisis is close to total in the areas that matter, and a site that cites a collapsed study on a health page has failed twice over.

In effect

Four notes that were wrong where being wrong costs most

The instructive case is this vault's own. Four notes carried the ymyl tag while still sitting in the unreviewed queue, which is the worst available combination: the material where an error does the most damage, in the pile nobody had checked. All four were verified on 1 September 2026, and three needed correcting.

The serious one was the note on attachment styles. It had silently merged two different models. Mary Ainsworth's Strange Situation, in 1978, identified three infant patterns; the fourth, disorganised, is Main and Solomon in 1990. The four adult styles the note actually went on to describe, secure, anxious-preoccupied, dismissive-avoidant and fearful-avoidant, are Bartholomew and Horowitz's 1991 two-dimensional model in the Journal of Personality and Social Psychology, and fearful-avoidant is routinely equated with disorganised although they are not the same thing. The note's claim that these patterns are laid down in the first two years and show up reliably in adulthood does not survive Fraley's 2002 review in Personality and Social Psychology Review, which finds modest continuity and explicitly models a revisionist account against a prototype one. Groh and colleagues relate infant attachment to later peer competence at about r = .19 and to internalising symptoms at about r = -.07. The sentence "one of the most clinically-validated frameworks in modern psychology" was deleted.

A second note, on motivational interviewing, had been credited to Adam Alter, who reports it in Irresistible and did not originate it. It is Miller in 1983, in Behavioural Psychotherapy, elaborated with Rollnick in 1991, and its efficacy is real but domain-dependent, strongest in alcohol and drug use.

The YMYL note itself was the only one in either tranche that needed no correction.

A reader typing a symptom into a search box

The everyday version has no vocabulary attached to it. Somebody types a symptom, a diagnosis or a drug name into a search box, usually at night, usually because they are worried, and they read the first three results with more credulity than they would give the same claims from a stranger in a pub. Every check a careful reader would run in that state is a YMYL check: who wrote this, what do they know, is there a study behind this sentence, and does this page have anything to lose if it is wrong. The classification exists because that reader has fewer defences than usual at exactly the moment the stakes are highest.

Three pillars out of nine

The commercial consequence for this publication was a scoping decision made during planning. Of nine pillars, three are YMYL territory: Why We Care, covering mental health, attachment and trauma; Why We Move, covering training and injury risk; and Why We Crave, covering eating, diet and eating disorders. Those three carry the full apparatus, primary-source citations, credentialed reviewers, disclaimer blocks, and the standing exclusion of the heaviest subjects. The other six are mostly not YMYL, being applied behavioural psychology for marketers and curious professionals, where the stakes are commercial rather than medical. The economics of that split are unattractive and worth naming: the three most expensive pillars to produce are also, in this field, three of the most searched.

What it does not say

It does not say YMYL pages rank differently. There is no flag, no score and no lever. The category tells human raters how strictly to judge, and their judgments feed back into evaluating algorithm changes rather than into the ranking of the pages they read.

It does not define its own boundary precisely. Where a topic stops being YMYL is a judgment, and the guidelines describe rather than enumerate. Sleep, exercise, relationships and money habits all sit in a band where reasonable editors would draw the line in different places, and the September 2025 clarification exists because the boundary needed restating.

It does not certify anything about a page that clears it. A page can meet every YMYL expectation on the site's part, named credentialed author, primary citations, disclaimer, review, and still repeat a finding that will not survive the next decade. The classification is about care taken, not about truth achieved. That gap is why this site publishes an evidence status on every concept page rather than relying on the appearance of rigour.

It is not a finding, and nothing here replicated. It is a company's published classification. Cite it as what it is, and check the date of whatever you read about it, because most of the writing on the subject describes a version that has since been clarified.

Who developed it

Google, in its Search Quality Rater Guidelines. The concept sits in that documentation from 2013 and was published openly in 2014, with the definitions clarified again in September 2025. There is no individual originator and no academic paper behind it. The vault's attribution was already correct when it was checked on 1 September 2026, which made this the only note in either verification tranche that needed no fix, and worth recording given how many neighbouring notes did.

---