Google pays external raters to score search results, and it publishes the manual it gives them. That document, the Search Quality Rater Guidelines, runs to 182 pages in the version currently in force, dated 11 September 2025. It does not describe the ranking algorithm. It describes the observation method Google asks human readers to apply.
This article takes the document from one angle: the checklist that applies to a whole site, the one you can borrow to re-read your own. Purpose of the page, how it is divided up, who the business is, reputation away from the site, state of upkeep.
The detail of E-E-A-T and the finer points of writing sit outside that scope; they belong with SEO content. What follows is the structural read.
What raters do, and what they do not
The manual says it in its opening section: no individual rating can directly influence how a page, a site or a result appears in Google, and none of them moves anything up or down. Google adds that ordering results from human ratings would not be possible at the scale of the web. The ratings measure whether the ranking systems are doing their job, and supply examples of helpful and unhelpful results.
Google's "Rigorous testing" page repeats the formula: these evaluations have no direct effect on ranking, they act as a quality benchmark. It also puts a number on the exercise: 719,326 search quality tests in 2023, a figure Google published for that year and still displayed in September 2026.
The practical consequence is worth stating plainly: you cannot please a rater, and nobody can sell you that. What you can do is borrow the checklist and read your own site the way a stranger would.
Start with the purpose of the page
The Page Quality rating asks one question first: how far does this page achieve its purpose? So the purpose has to be identified before anything else. The manual insists on a point that is often misread. No type of page is inherently better than another. An encyclopaedic page is not automatically better than a humorous one, as long as each does what it sets out to do.
What varies is the level of expectation. The rating considerations separate small amateur sites from large business sites, and sites that handle financial transactions from those that ask for neither payment nor personal data. A family blog judged on its ability to share news can be excellent; a shop assessed with the same indulgence would not be.
YMYL describes a topic, not a sector
YMYL, Your Money or Your Life, is regularly presented as a category of sites: those dealing with money, health or personal data. The 2025 version says something different. YMYL describes topics that could significantly affect people's health, financial stability or safety, or the welfare of society. Four families are listed: health or safety, financial security, government and society, and an "other" category.
The assessment runs along a continuum: clearly YMYL, possibly YMYL, not YMYL at all, with most topics in the last group. For a small business the question therefore arises page by page. A heating engineer explaining gas safety rules is writing on a sensitive topic; the same engineer publishing opening hours is not. The first kind of page is where accuracy and a named author have to be beyond reproach.
Main content, supplementary content, advertising
The guidelines cut every page into three blocks. The split is useful because it forces you to look at a page as a visitor rather than as its publisher.
| Block | Google's definition | What pushes the rating down |
|---|---|---|
| Main content (MC) | Everything that directly helps the page achieve its purpose: text, images, video, features. The title displayed at the top of the page is part of it. | Main content missing, too thin, or deliberately obscured by interstitials and advertising. |
| Supplementary content (SC) | What contributes to the experience without directly serving the purpose: navigation, related links, side panels. | Elements distracting enough to get in the way of the information the visitor came for. |
| Advertising and monetisation | Any content or link displayed to monetise the page, affiliate links included. | Nothing in itself: the presence or absence of advertising is not a reason for a high or a low rating. |
Google is explicit on that last point: without monetisation, some pages would not exist at all. What gets penalised is obstruction. Adverts that cover the text as you scroll, a window that cannot be closed without a click, content pushed so far down that many visitors give up: those cases fall into the lowest rating. Read that column with your own consent banner, newsletter pop-up and sticky header in mind.
Who is behind the site?
Two whole sections are given over to an ordinary question: can a reader find out who is responsible for the site, who wrote the page, and how to reach the business? Raters are asked to look for an about page, a contact page, and, for a shop, customer service information, including payment, exchange and returns terms.
The expected level depends on what the site does. A personal site can stay discreet about its contact details. A page that takes a payment should be rated low if contact or customer service information is insufficient. A page on a YMYL topic, or a site handling sensitive data, with nothing at all about who is responsible for it, falls into the lowest rating.
These expectations overlap with the disclosure duties that apply to businesses trading online in most jurisdictions, so check what your own law requires. Meeting them earns no ranking bonus, but neglecting them stacks up negative signals for readers and raters alike, and it is among the cheapest things to fix during a UX audit.
Reputation is checked somewhere other than your site
Reputation research is mandatory for every page quality rating task, and it happens off the site. The manual gives the query patterns, along the lines of [ibm -site:ibm.com], and points out that a social media page run by the business itself is not an independent source. Raters look for articles and expert recommendations written by third parties.
Customer reviews count, with explicit reservations. Raters are asked to stay sceptical of positive and negative reviews alike, to look for a large number of them, and above all to read them. Credible reports of fraud amount to a very negative reputation; a late delivery does not.
One last point, reassuring for a small organisation: an absence of reputation information is neither a positive nor a negative signal. The manual notes that plenty of local businesses run on word of mouth. The other criteria then carry more weight, and the practical work moves towards directories, map listings and review platforms, the territory of local SEO.
A neglected site, a hacked site
Contrary to a widespread belief, freshness is not a central criterion in the quality rating. An old page can score highly, and some reference publications maintain archives on purpose. What the guidelines target is an abandoned site, or content that has become inaccurate because nobody revised it. The distinction matters when a maintenance budget is allocated: republishing dates achieves nothing, correcting what is wrong achieves a great deal.
Hacking is treated without nuance. A page that has been hacked, defaced or buried under spam should receive the lowest rating, because it no longer achieves its original purpose. Google adds that a responsible owner checks regularly for suspicious behaviour. If unfamiliar pages start appearing in your search results, our page on a hacked WordPress site sets out the sequence to follow.
Using the checklist without over-reading it
Take five questions, in the order the manual puts them:
- Is the purpose of the page obvious within a few seconds?
- Does the topic touch on health, money or safety?
- Is the main content immediately accessible, or buried under other things?
- Can a reader tell who is responsible for the site and how to reach them?
- What does someone find about you elsewhere on the web?
A weak answer on any of those hinders a visitor first, and search performance second. That order tells you what to fix when the two appear to conflict.
Keep the limit in mind as well. The document describes the target Google is trying to hit, not the signals it computes. Following it guarantees no position. Ignoring it leaves defects that any attentive reader spots in three minutes.
Common questions
Can a Google rater push my site down?
No. The official guidelines state that no individual rating moves a page, a site or a result up or down in Google. Ratings measure the quality of the ranking systems and provide examples of good and poor results.
Where do I find the current version of the Search Quality Rater Guidelines?
Google publishes the PDF openly on its own distribution domain, in English. The version in force in September 2026 is dated 11 September 2025 and runs to 182 pages. The date appears at the top, and a change log sits in the appendix.
Does YMYL apply to my site?
YMYL describes topics, not business sectors. A site can have a handful of pages on a sensitive topic, such as health, money or safety, and many others that are not sensitive at all. Those few pages demand the most rigour on accuracy and on who the author is.
Should I remove advertising to be rated better?
No. The guidelines are clear that the presence or absence of advertising is not in itself a reason for a high or a low rating. The problem is advertising that covers the content, prevents it being read, or pushes it a long way down the page.
What if nothing about my business exists outside my own site?
That is not a handicap in itself. The manual says an absence of reputation information is neither a positive nor a negative signal. The other criteria then carry more weight: clarity of purpose, quality of the content, contact information.
Do I need to republish pages regularly to look maintained?
No. The guidelines note that an older page can receive a high rating and that freshness is not a central criterion of the quality rating. What is penalised is an abandoned site, or content that has become inaccurate because it was never revised.
Is this a substitute for a technical audit?
No. This grid covers what a human reader can assess. Crawling, indexing and speed need their own examination.