Skip to content
notes · earned authority

The quality raters don’t grade your site

Structure and markup refreshed for current answer engines; the original analysis is preserved.

Google publishes the Search Quality Rater Guidelines and employs thousands of human raters who score results from Lowest to Highest — but their scores do not touch your ranking. They calibrate the algorithm in aggregate; no rater grades your page. The real value of the guidelines is as the clearest public picture of what Google means by quality: E-A-T, beneficial purpose, YMYL. Read them to learn what to build.

the short answer

Google’s thousands of human quality raters score samples of results from Lowest to Highest, but their scores do not affect your ranking directly — there is no rater score on your page. They calibrate the algorithm in aggregate; the algorithm then ranks everyone. The guidelines’ real value is as a public specification of what Google means by quality: E-A-T (Expertise, Authoritativeness, Trustworthiness), beneficial purpose, and the higher bar for YMYL topics. Read them to learn what to build, not as a scorecard for your site.

key takeaways

  • Google publishes the Search Quality Rater Guidelines and employs thousands of human raters who score results from Lowest to Highest.
  • Their scores do not touch your ranking. Raters calibrate the algorithm in aggregate; no rater grades your individual page or leaves a score on it.
  • The real value of the guidelines is as the clearest public picture of what Google means by quality — E-A-T, beneficial purpose, YMYL, reputation.
  • E-A-T (Expertise, Authoritativeness, Trustworthiness) is assessed partly from outside your site, through reputation research. It is accumulated, not set.
  • Read the guidelines as a specification of the target the algorithm is built to approximate, and build toward it — not toward an evaluator who was never assigned to you.

the myth vs the reality

the myth · a human grades you rater → looks at your site → leaves a score on your page → moves your ranking ✗ no rater is assigned    to your URL the reality · raters calibrate raters score samples (Lowest→Highest) → aggregate feedback → calibrates the algorithm → algorithm ranks everyone ✓ guidelines = public spec:    E-A-T · purpose · YMYL No one grades your site by hand. Read the guidelines to learn what the algorithm is built to reward.

The left side imagines a grader who was never assigned to you; the right side is the actual chain — humans tune the machine, the machine ranks the web. Once you see the two steps as separate, the guidelines stop being a verdict to fear and become a specification to use.

The argument, in four parts

The misread is that a human grades your page; the reality is that raters calibrate the algorithm in aggregate; the guidelines are worth reading as a specification; and that spec keeps asking for earned, external quality. Open each part.

01 The misread: a human is grading your page

The most common misunderstanding about the quality raters is also the most natural one: that somewhere in Google there is a person who looked at your site, formed a verdict, and nudged your ranking up or down accordingly. It is easy to believe because the raters are real — thousands of them, working from a long, public rulebook, assigning scores from Lowest to Highest — and because the existence of human judgement seems to imply human judgement of you. If people are paid to look at search results and decide whether they are good, surely one of them looked at yours and decided. The inference feels airtight, and it is wrong at the first step, because the unit they evaluate is the result set, not the site, and the conclusion they reach is about the system, not the page. But that is not how the system is wired. The raters never score your site as your site; they score samples of search results as evidence about whether the ranking systems are working. The verdict they produce is about the algorithm, not about you, and it is delivered in aggregate, far upstream of any individual page — pooled with thousands of other judgements into a number that describes how well the system is doing, never a number that describes how well your site is doing. By the time that number reaches an engineer, it has no memory of any single result, let alone any single domain; it has been averaged into a verdict about the system that could not name your site if it tried. That is not a loophole or an oversight — it is the design, because the whole point of the exercise is to judge the engine, and judging individual sites by hand would not scale to a web this size even if Google wanted to. There are hundreds of millions of sites and a finite number of raters; the arithmetic alone rules out a per-site grade, and the system was built around that fact from the start rather than in spite of it. Picturing a grader hovering over your URL is picturing a relationship that does not exist — a named person with your domain open in a tab, deciding your fate. The raters have never seen your site as a site to be ranked; they have only ever seen results to be scored, and the difference between those two things is the whole point.

02 How the raters actually fit in

What the raters do is closer to quality control on a production line than to grading a paper. Google makes a change to its ranking systems, or wants to know how well the current systems perform, and sends out tasks: here is a query, here are some results, score them against the guidelines, and move on to the next. The rater is not investigating a business; they are sampling the engine’s output and reporting how well it matched the standard, one result set at a time. Many raters score many samples across many queries, and the aggregate tells Google whether a candidate change made results better or worse before it ever ships, and where the live systems are falling short of what the guidelines describe. A single rater’s opinion of a single result carries almost no weight on its own; it is the pattern across thousands of independent judgements that becomes a usable signal about the overall health of the system, much the way a public-opinion poll draws all of its meaning from the shape of the whole sample rather than from any one respondent in it. That feedback shapes the algorithm, which then ranks everyone — including you — by signals it can compute at scale. Your page is affected only at that second remove, through an algorithm the rater feedback helped tune, never through a rater touching your page. It is the difference between a chef tasting a sauce in the kitchen to adjust the recipe and a chef coming out to taste the specific plate in front of you: the first quietly improves what everyone is served from then on, the second never happens at all. No one is coming out to taste your plate, and waiting for that visit — or dreading it, or trying to dress the plate for it — is spending energy on an event the kitchen never scheduled and never will. The raters taste the recipe, and the recipe is what reaches your table. The humans calibrate the machine; the machine ranks the web. Collapsing those two steps into one is the root of the misunderstanding, and it is an easy collapse to make, because from the outside the only visible facts are that humans evaluate quality and that rankings change. The mind fills the gap between them with a story — the human must have changed the ranking — and the story is wrong in a way that quietly misdirects everything you do next.

03 Why the guidelines are worth reading anyway

None of this makes the guidelines irrelevant to you — it makes them valuable in a different way than people assume. Because the algorithm is built to approximate the same quality the raters are told to score, the rulebook is effectively a public specification of the target: it describes, in Google’s own words, what a high-quality, trustworthy, useful page is supposed to be. That is rare and worth using. Instead of inferring what Google wants from the noise of ranking changes, you can read what raters are instructed to reward and to penalise, and build toward the former. The document will not tell you the algorithm’s weights, but it will tell you its intent, and intent is enough to direct real work. Read as a spec rather than a scorecard, it stops being a source of anxiety and becomes a source of direction. The shift is not just psychological; it changes the questions you ask. Instead of “what did a rater think of my page,” which has no answer, you ask “what does the standard reward, and where do I fall short of it,” which has a long and actionable one.

04 What the spec actually asks for

Read that way, the guidelines keep returning to a few things. They ask raters to judge expertise, authoritativeness, and trustworthiness — E-A-T — not only from the page itself but from reputation research: who created this, and how are they regarded elsewhere, in reviews, references, and independent coverage. They ask whether the page has a beneficial purpose and achieves it well, treating pages built to help people very differently from pages built mainly to extract from them — to capture a click, serve an ad, or harvest a lead while the promised answer stays thin or buried. That distinction runs through the whole document, and it is one of the clearest signals of what the algorithm is ultimately reaching for. And they single out Your-Money-or-Your-Life topics — health, finance, safety, civics, and the like — for the highest scrutiny, because the cost of being wrong there is real, measured in someone’s health, money, or safety rather than in a worse afternoon. That is why a thin, anonymous page on a casual hobby might pass without much scrutiny while the same thinness on a medical or financial topic is treated as a genuine failure: the guidelines scale the demand for expertise and trust to the stakes of the subject, and so, by design, does the algorithm built to approximate them. The throughline is that quality, as Google describes it, is largely earned and largely external: it lives in who vouches for you and how well you serve a genuine need, not in anything you can assert on your own page. That is an uncomfortable truth for anyone hoping for a checklist of on-page tweaks, but it is also a liberating one, because it means the work that matters is the work of actually becoming good and known for it — not the work of decorating a page to look that way to a grader who is not coming.

Why the distinction changes what you do

Believing a rater grades your site sends you looking for the wrong remedy. If you think a human verdict is sitting on your page, you start trying to appeal it — tweaking surface signals, hunting for the specific thing that offended an evaluator, treating ranking movement as a grade you can argue with. None of that connects to anything real, because there is no verdict and no evaluator to appeal to. The version that is true points you somewhere more productive: the algorithm is approximating a public standard, the standard is written down, and your job is to meet the standard rather than to satisfy a phantom grader. That reframing turns a source of anxiety into a piece of reference material, and it redirects effort from chasing fluctuations toward building the thing the guidelines actually describe. The team that internalises it stops refreshing the rank tracker for clues and starts reading the standard for requirements, which is a quieter and far more productive way to spend the same hours.

And what they describe is largely earned and largely external, which is the part most worth internalising. The guidelines push raters to research who stands behind a page and how that source is regarded elsewhere, to ask whether a page genuinely helps the person who lands on it, and to hold the highest bar for the topics where being wrong does real harm. You cannot set any of that from inside your own markup; you accumulate it — through expertise that shows, a reputation others confirm, and a purpose that serves a real need. That is exactly the kind of authority the AC Group has spent 27 years helping organisations build, because it is the kind that an algorithm tuned to a public standard is designed to find, and the kind no phantom grader and no quick fix can manufacture for you, because it is made of things that take time — demonstrated knowledge, accumulated reputation, a record of having served a real need well — and that resist being faked precisely because the standard was written, in part, to help a careful human see through the fake. Anything quick enough to fabricate in an afternoon is, almost by definition, the kind of thing the guidelines teach a rater to discount, which is why the durable move is to stop looking for the shortcut and start accumulating the real thing.

What to do with this

Read the guidelines once, as a specification rather than a scorecard, and let them set your priorities. Note where they ask for expertise, authoritativeness, and trustworthiness, and be honest about where your pages and their authors fall short of being a recognised, well-regarded source. Note where they describe a beneficial purpose, and check whether your most important pages are built to help the visitor or to extract from them, because the guidelines draw that line sharply and the algorithm is built to find it. A page that answers the question it promised to answer reads very differently from one that buries the answer beneath obstacles designed to detain the reader, and the difference is exactly the kind of thing the standard is written to catch. If your topics fall inside Your-Money-or-Your-Life — and for most commercial sites some of them will — accept that the bar is higher and that thin, anonymous content is explicitly disfavoured, named in the guidelines as the kind of thing a rater should mark down rather than overlook. For a commercial site that is not a warning to soften; it is a requirement to meet. Then build toward the standard: clearer authorship, verifiable expertise, a reputation you can point to, and pages that earn the trust the guidelines keep asking for.

What you should not do is treat any of this as a trick to reverse-engineer or a grade to contest. There is no rater to satisfy and no score to flip; there is only the slow, durable work of becoming the kind of source the standard rewards. It is tempting to want the other thing — a lever, a setting, a fix that pays off this week — but the thing that actually moves you is also the thing that takes the longest, and pretending otherwise just postpones the real work while the quarter slips by. That is less immediate than tweaking a signal and far more reliable, because it aligns you with what the algorithm is built to approximate rather than with a guess about its mechanics. Building that earned, externally-confirmed authority — and reading Google’s own description of quality as the specification it is — is the steady approach the AC Group has worked by for 27 years.

Quality raters, plainly: quick answers

Do Google’s quality raters affect my ranking?

Not directly, and not for your individual page. Google has been explicit that the scores raters assign do not change the ranking of the sites they look at. Raters work on samples of search results, score them against a public rulebook, and that feedback is used in aggregate to measure and calibrate the ranking systems — to tell Google whether an algorithm change made results better or worse. There is no rater score attached to your URL, no human verdict sitting on your page waiting to be appealed. If your rankings move, it is the algorithm responding to signals, not a rater having visited and judged you. Treating the raters as graders of your specific site is chasing an evaluator who was never assigned to you.

Then why should I care about the rater guidelines at all?

Because they are the clearest public description of what Google means by a high-quality page, and the algorithm is built to approximate the same thing the raters score by hand. The guidelines spell out how raters judge expertise, authoritativeness, and trustworthiness; what a beneficial purpose looks like; why Your-Money-or-Your-Life topics are held to a higher bar; and how reputation is researched. None of that is the ranking algorithm itself, but all of it is the target the algorithm is trying to hit. So the document functions as a specification: it tells you, in Google’s own words, what "good" is supposed to look like, which is far more useful than guessing from ranking fluctuations.

What does E-A-T mean in the guidelines?

E-A-T stands for Expertise, Authoritativeness, and Trustworthiness — the three qualities raters are asked to weigh when judging the quality of a page and its source. Expertise is about the creator genuinely knowing the subject, in a way that shows in the substance rather than the styling; authoritativeness is about the site or author being a recognised, go-to source for it, the name others in the field would point to without having to think; trustworthiness is about the page and its publisher being reliable, transparent about who they are, and safe to act on — and of the three, the guidelines treat this last as the one that can sink a page on its own, since a reader stakes something real on it; trustworthiness is about the page and publisher being reliable and safe to act on, the quality the guidelines treat as most important of the three because it is the one a user is staking something on when they follow the advice, buy the product, or trust the claim. Expertise without trust is a credential no one can rely on; the guidelines want both, in that order of consequence, and they are unusually willing to say that a page failing on trust should be rated poorly no matter how expert or polished it otherwise appears. It is the one shortfall the document treats as close to disqualifying on its own, which tells you how much weight the underlying systems are likely to place there too. A page can be expert and authoritative and still fail on trust if it hides who is behind it or pursues a purpose other than helping the reader, and the guidelines are unusually direct that such a page should not score well no matter how polished it looks. The guidelines tell raters to research who is behind the content and how they are spoken about elsewhere, which means E-A-T is assessed partly from outside your site, not just from what you say about yourself. It is not a score you can set; it is a reputation you accumulate.

What is YMYL, and why does it get extra scrutiny?

YMYL stands for Your Money or Your Life: topics where bad information could harm someone’s health, finances, safety, or wellbeing — the current guidelines list areas such as finance, health and safety, civics and law, shopping, and news among them. Raters are told to apply the highest page-quality standards to these pages, because a wrong answer in these areas carries real consequences, and Google’s rankings function as a kind of endorsement. For most businesses this matters because commercial and transactional pages often fall inside YMYL, which means the expertise and trust bar for them is higher than for a casual topic. If you operate in one of these areas, the guidelines are telling you plainly that thin or anonymous content will not be treated kindly.

A note on sources and timing

This is written in March 2022. We have described the Search Quality Rater Guidelines as Google publishes them — a public rulebook for thousands of human raters who score result samples from Lowest to Highest, whose scores do not directly affect any site’s ranking but instead calibrate the systems in aggregate. We have used the framework as the current guidelines do: E-A-T (Expertise, Authoritativeness, Trustworthiness), beneficial purpose, and the higher bar for YMYL topics. We have framed the document as a specification to build toward rather than a scorecard to contest. The durable point holds regardless of the next revision: quality, as Google describes it, is earned and external — the kind of authority the AC Group has built for 27 years.

Measured against the standard, how do you score?

Our free audit reads your most important pages the way the guidelines ask a rater to — expertise, authority, trustworthiness, beneficial purpose, and the reputation that backs them — and tells you where you meet the bar and where you fall short. In English and Spanish, in 48 hours, with no sales call.