The search quality rater guidelines
These guidelines address human raters, not an algorithm. The document writes that no rating can move a page in the results: the ratings serve to measure and improve the engine. They therefore describe an ideal of quality, never a scale you could aim at.
What is this document made of?
A working manual for people paid to look at pages and rate them. It has a sober title, General Guidelines, it is dated 11 September 2025, it runs to a hundred and eighty-two pages, and it is published openly by the engine’s vendor.
It is neither technical documentation nor the description of an algorithm. It is what you hand to someone about to evaluate search results and say whether they are good. That distinction is not a subtlety: it decides every use the document can be put to.
What the document says about itself, on its second page
No rating can directly influence how a page, a site or a result appears in search, nor move a result up or down the page; the document states that positioning results that way would not be feasible, since humans cannot individually rate every page on the open web.
officialSource : Google, “General Guidelines”, version of 11 September 2025
The sentence sits in the section explaining to raters what their work is for, before any instruction. It comes with a practical reason almost never quoted: humans could not individually rate every page on the open web. That is not legal caution, it is a constraint of scale.
So what are the ratings for?
The ratings serve to measure how well the engine delivers useful content, and to improve it by supplying examples of useful and unhelpful results.
officialSource : Google, “General Guidelines”, version of 11 September 2025
In other words: they measure the engine, not the sites. A rater judging a page poor produces a data point about the quality of results, and possibly an example that will later serve to assess a change. They punish nothing and reward nobody.
Lost positions? A Google quality rater has probably marked your site “low quality”. Our E-E-A-T audit identifies the criteria you were penalised on and gets your positions back.
E-E-A-T, and what those four letters cover
The document devotes a section to four notions — experience, expertise, authoritativeness, trust — grouped under the initials E-E-A-T. Of the four, one counts more than the others, and the text says so plainly.
Trust is the most important member of the E-E-A-T family, because a page that does not inspire trust has low E-E-A-T whatever the apparent experience, expertise or authoritativeness of its author.
officialSource : Google, “General Guidelines”, version of 11 September 2025
The example given in the document has stayed famous for its bluntness: a financial scam does not inspire trust, even written by a very experienced, expert and well-known scammer. The first three letters can be at maximum without the fourth following, and it is the fourth that decides.
What the document does not do, on the other hand, is assign a value.
E-E-A-T is a vocabulary for describing what a human rater observes on a page, not a value a page would carry and a tool could read off.
hypothesis
That status is deliberate. The document contains no sentence saying “E-E-A-T is not a score”, and we are not going to put one in its mouth. What can be observed is that it describes no scale, no calculation and no transferable value, and that it addresses human readers asked for a reasoned judgement. The rest is a reading, displayed as one.
The topics where the bar rises
Topics called “your money or your life” are those whose treatment could have a significant effect on people's health, financial stability or safety, or on the welfare of society.
officialSource : Google, “General Guidelines”, version of 11 September 2025
The document separates two families inside that category, and the distinction is useful. There are topics dangerous by nature, and there are those that only become dangerous if the content is inaccurate — the symptoms of a heart attack, where to place a sum of money, what to do during an earthquake. In the second case it is not the topic that harms, it is the error.
A site handling those topics is subject to no special ranking rule. It is simply evaluated, by the people charged with measuring the quality of the engine, with a higher bar on accuracy and on identifying who is writing.
What the guidelines are worth anyway
Everything above might suggest this document is useless to a site publisher. The opposite is true, provided it is read for what it is.
It describes, over a hundred and eighty-two pages and with examples, what the engine’s vendor considers a trustworthy page: who writes, how you know, what the site says about itself, what is found about it elsewhere, what becomes of the main content when advertising covers it. No other public source is as detailed on that question.
Ratings must not rest on the rater's personal opinions, preferences, religious beliefs or political views, but on the instructions and examples in the document.
officialSource : Google, “General Guidelines”, version of 11 September 2025
That instruction deserves keeping by anyone who writes. It means the document is after a judgement reproducible from one rater to another, founded on clues observable on the page and around it. That is exactly what a demanding reviewer can apply to their own texts, without ever claiming to reproduce a score.
Rereading your pages with the rater guidelines as a grid
hypothesis
- Effort
- high
- Time to effect
- quarters
- Risk
- none
- Durability
- structural
The effort is real: the document is long, and a useful reading means skipping the half devoted to the mechanics of rating tasks. The time-to-effect axis is the longest on this site, because no effect is expected from the rereading itself — it changes what gets published, and it is the content that works afterwards.
The risk is as low as it goes: nothing this document describes is a technique, still less a risky one. And durability is good, which can be checked rather than trusted: the document has been reissued regularly under the same title for years, with a version number and a date, which makes it possible to compare editions instead of believing them.
What does not work, or no longer works
Looking in this document for a list of criteria to tick, first. It contains none; it contains annotated examples, which is exactly the opposite of a scale.
Then, dating your pages, displaying an invented author and adding an empty “about” page in the hope of ticking a box. The document specifically asks that what is found elsewhere than on the site about its author be sought out — reputation has its own section there, and it is not declared about oneself.
Finally, treating each new edition as a revelation. The changes from one version to the next are readable, often modest, and one dated monitoring entry is worth more than the twentieth analysis of a four-letter acronym.
Sources
- General Guidelines, version of 11 September 2025, 182 pages — Googleaccessed on
- Creating helpful, reliable, people-first content — Google Search Centralaccessed on
Block revalidated on
Frequently asked questions
- Can raters move my site up or down?
- No, and the document says so in its introduction: no rating can move a result on the page. It adds a practical reason rarely quoted: rating every page on the open web one by one would not be feasible. The ratings measure the engine, they do not rank sites.
- Is there an E-E-A-T score you can measure?
- No tool reads such a value, and the document describes none. E-E-A-T there is an observational vocabulary aimed at human readers. A supplier selling a score out of a hundred is selling their own scale, which may be useful as a review grid but has nothing to do with the engine.
- Should the guidelines then be ignored?
- No. They describe precisely what the vendor considers a trustworthy page, and as such they are the best public review grid available. What must not be done is treating them as a ranking scale, because they say themselves that they are not one.
- What is a “your money or your life” topic?
- A topic whose treatment could weigh on people's health, financial stability or safety, or on the welfare of society. The document separates topics dangerous by nature from those that only become dangerous when inaccurate, such as heart attack symptoms or where to invest money.