Request analysis →
FR EN
Algorithmic analysis of EEAT signals
№70 Qr Core
Core · E-E-A-T pillar

Quality Rater Guidelines: Google's manual, explained

📖 13 min min read · Cross-cutting concept · the grid Google uses to define quality

Why this signal matters for Google

The quality rater guidelines are the only document where Google writes down what it means by quality. This is not a leak, and not a blogger's reconstruction. It is a public manual, republished regularly, and the version in force is dated September 11, 2025. Google publishes it because it has to. Without a written, shared definition, thousands of outside evaluators would each rate to their own taste. The document exists so that quality always gets judged the same way. And nothing stops you from using it exactly the way they do.

An evaluator does not score your position in Google. They give two different ratings. Part 1 covers Page Quality, rated from Lowest to Highest, with no search involved. Part 3 covers Needs Met (section 13.0): it judges the search and the result together, from Fails to Meet to Fully Meets. Mixing up the two is the mistake I run into most often in audits. An excellent page deserves Fails to Meet on a search it does not answer. And these ratings change nothing about a page's position · they measure the quality of the results and train the systems that do the ranking.

E-E-A-T is their working tool before it is an SEO concept. It appears in section 3.4 as a component of page quality, with trust at the center: an untrustworthy page does not earn a good rating, even when a well-known expert signed it. Section 3.4.1 raises the bar as soon as the topic is YMYL, meaning health, money or safety. Section 3.2 ties the quality of the main content to how accurate it is. That order · trust first, expertise second · is exactly the one most SEO audits reverse, by looking at the author's credentials before looking at whether the page holds up.

Thinking like an evaluator is also the best preparation for AI answers. No public documentation says AI Overviews apply the Page Quality scale, and claiming otherwise would be dishonest. But the question an AI engine has to settle before citing a source is the evaluator's question: can this page be trusted on this topic, and does it really answer the request? The pages I see cited in ChatGPT and Perplexity answers are almost always the ones that would clear an evaluator's review without argument. The habit behind it is a dull one: every claim tied to its cited sources, so an evaluator and a model can both check it in one click.

The NeuroEEAT scoring grid (0-10)

This scale does not rate a page. It rates the way you evaluate: how closely it matches a Google evaluator's. Aim for 7/10 if your site touches YMYL topics.

ScoreDescriptionConcrete indicators
0-2Guidelines unreadThe document has never been opened. Quality is judged on gut feel, or on a score from an outside tool that has nothing to do with the criteria Google writes down.
3-4Secondhand knowledgeE-E-A-T comes from blog posts rather than section 3.4. Page quality and needs met are still mixed into one vague idea.
5-6Read, never appliedThe version in force has been skimmed and the useful sections spotted, but no page has ever received a written rating.
7-8Self-assessment runningKey pages carry two written ratings, page quality and needs met. The page purpose and the YMYL status are settled before any judgment.
9-10Repeatable protocolThe assessment is written down, redone by two people who do not compare notes, replayed after every rebuild, and the rating gaps decide the editorial calendar.

How to audit it on your site

Auditing this concept is not about rating one page: it is about checking that your way of rating holds up. Five steps, in the order an evaluator follows them.

  1. 01

    Work from the version in force

    Download the official quality rater guidelines PDF from Google and check the date: the version in force is dated September 11, 2025. A blog post built on the 2019 edition describes a grid where Experience does not exist yet, since it was added in December 2022.

  2. 02

    Write the page purpose before judging it

    An evaluator starts by working out what the page is trying to do. A page with no clear useful purpose cannot earn a good rating, however well written it is. Write that purpose in one sentence first: the exercise alone flushes out pages written for a keyword rather than a reader.

  3. 03

    Place the topic on the YMYL scale

    Section 2.3 sets how strict everything that follows will be. A clearly YMYL page and a page that is not are not rated the same way. Skip this step and every rating becomes unusable: far too harsh on a hobby post, far too generous on a financial advice page.

  4. 04

    Give a page quality rating

    Rate from Lowest to Highest by looking at the accuracy of the main content (section 3.2), the reputation of the site and its author, and the E-E-A-T of section 3.4. An untrustworthy page hits a ceiling however cleanly it is written · the constraint surface-level audits never see.

  5. 05

    Give a needs met rating, search by search

    Start from the search, not the page (section 13.0). For your ten most important searches, ask whether your page earns Fully Meets or only Slightly Meets. The gap between that rating and the page quality rating tells you whether to rewrite the page or target a different search.

How to activate or improve it

Three levels to move from quality judged on gut feel to an assessment anyone can repeat the same way.

Beginner

Read the grid and apply it once

  • Download the current official PDF and read Part 1 end to end
  • Write the real purpose of your five most visited pages in one sentence each
  • Place each of them on the YMYL spectrum before judging them
  • Give each page a written rating from Lowest to Highest
Intermediate

Split the two ratings

  • Add a needs met rating per target search, separate from the page rating
  • For every low rating, note the section of the guidelines that justifies it
  • Have the same pages rated by someone who did not write the content
  • Rewrite first where the gap between the two ratings is widest
  • Redo the exercise after every rebuild or major update
Pro

Industrialize the rating

  • Build a scorecard that follows the guidelines order: purpose, YMYL, page quality, needs met
  • Train writers on the section 3.4 order: trust before expertise
  • Add the rating to the list of boxes to tick before publishing
  • Reread the guidelines with each new version and update the scorecard
  • Compare your ratings against real results to adjust how strict you are

5 common mistakes to avoid

❌ Believing an evaluator scores your site and can punish it · a low rating triggers no action against an address
✅ Treating the grid as a self-assessment tool · you rate your pages before the systems trained on those ratings do it for you
❌ Mixing page quality and needs met into one judgment · one rating for two different questions
✅ Two separate ratings · quality with no search involved, needs met for each target search (section 13.0)
❌ Quoting an outdated edition · summaries of the 2019 version describe a grid with no Experience
✅ Working from the official PDF in force · the September 11, 2025 version, downloaded from Google directly
❌ Rating the author's expertise first · that is the reverse of the guidelines order
✅ Rating trust first · section 3.4 puts it at the center, and expertise does not redeem an untrustworthy page
❌ Hunting for an official numeric score in the guidelines · evaluators assign no rating out of 10 or out of 100
✅ Using the real scales · Lowest to Highest for the page, Fails to Meet to Fully Meets for intent, and an owned 0-10 scale purely as an internal tracking tool

Sector specifics

YMYL · Health / Finance

Maximum strictness

Section 3.4.1 raises the bar the moment a page informs or advises on health, money or law. Rate these pages assuming an evaluator reads them asking whether they could do harm: an identifiable party responsible for the site, an author qualified for the advice given, facts in line with what the medical or financial field actually says. An average page anywhere else is a weak page here.

Local · Neighborhood business

Needs met leads

On a local search, practical information makes the rating: address, hours, contact method, service area. An accurate, up-to-date 300-word page deserves better than a long one-size-fits-all article · which is exactly what the needs met rating in section 13.0 measures.

Publishing · Media

Reputation moves to the center

On a content site, the evaluator looks at the reputation of the site and its authors as much as the text. A published editorial policy, dated corrections and consistent bylines then weigh as much as the craft of any single article.

E-commerce · Transactional

Operational trust

A page that takes a payment is rated on concrete trust: customer service that is easy to find, clear return terms, an identifiable party responsible for the site. Missing contact information alone caps the rating here, however good the product pages are.

Signals to activate in parallel

Quality Rater is not a signal you activate: it governs how you rate every other one. These five signals are the ones it shapes most directly:

FAQ

Where can I find the quality rater guidelines, and in what format?

Google publishes the document as a PDF and updates it regularly: the version in force is dated September 11, 2025. Look for it on Google's own site rather than a copy hosted elsewhere, because those copies tend to sit on outdated editions. A 2019 file describes a grid where Experience does not exist yet.

Can a quality rater penalize my site?

No. An evaluator never touches a page's position. Their ratings measure the quality of search results and train or check the systems that do rank. A poor rating therefore triggers nothing against your address, but it describes very precisely what those systems are trying to reproduce at scale.

Is E-E-A-T a ranking factor?

Not in the sense of a dial inside the algorithm. It is a human evaluation grid described in section 3.4 that Google uses to judge whether its own results are good. The difference matters in practice: you do not optimize an E-E-A-T score, you make visible the evidence an evaluator looks for.

Do I need to read the whole document?

Part 1 on page quality, yes, at least once: that is where E-E-A-T, reputation and YMYL live. Part 3 on needs met comes next, read with your list of searches in front of you. The rest mostly covers rating situations specific to the evaluators' own workflow and will do little for you.

How do I use the quality rater guidelines without being a rater?

By redoing their work on your pages: page purpose, YMYL status, page quality rating, needs met rating. Two hours spent that way on your ten main pages surfaces more problems than a full technical audit, because you stop judging your content by your criteria and start judging it by theirs.

Anonymized case study

Illustrative scenario · for illustration only, validate with your own data

Before · quality judged on instinct

Page purpose written downNever
YMYL status assessedNo
Page quality ratingNone
Needs met ratingNone
Signal score (indicative)2 / 10

After · rater protocol applied

Page purpose written downOn the 20 main pages
YMYL status assessedPlaced on the spectrum, page by page
Page quality ratingDocumented with the section that justifies it
Needs met ratingOne rating per target query
Signal score (indicative)9 / 10

Illustrative figures meant to show the typical progression of the evaluation practice. They do not come from a real client case and must be replaced with your own measurements before any external communication.