Google doesn't rank pages on looks. It ranks on trust. Behind every search result is a framework, applied by thousands of contracted quality raters, that defines what "good" looks like. Learn their rubric and you learn what Google's ranking systems are trying to reward.
What quality raters actually do#
Quality raters are not the algorithm. They rate sample search results using a public document called the Search Quality Rater Guidelines, and Google uses those ratings to measure how well its ranking systems work and to evaluate proposed changes before launch. In Google's own words, the guidelines are "not a guide to ranking first in Google" and raters' "ratings don't directly influence ranking". John Mueller put it the same way in May 2026: the guidelines are insightful, but not a ranking guide.
Translation: what raters are told to call "excellent" is the target Google tunes its systems toward. No rater can push your site up or down.
What changed in 2025#
- January 2025: the Lowest and Low page quality sections were rewritten to align with Google's spam policies. Trade press noted the edition added a definition of generative AI, rater guidance on expired domain abuse, site reputation abuse and scaled content abuse, and a Lowest rating when all or almost all main content is copied, paraphrased or AI-generated with little effort, originality or added value. Telltale AI phrases left in content ("As an AI language model...") are listed as a sign of low effort.
- September 2025: YMYL definitions were updated (see below), more examples were added, plus minor fixes.
The E-E-A-T framework#
E-E-A-T stands for Experience, Expertise, Authoritativeness, and Trust. Google added the first "E" for Experience in December 2022, a tacit acknowledgement that real first-hand knowledge beats polished-sounding generalism.
| Signal | The question it answers | What raters look for |
|---|---|---|
| Experience | Has this person actually done this? | First-person narrative, original photos, real data from real testing |
| Expertise | Do they know this subject deeply? | Credentials, depth of content, cited sources, no obvious errors |
| Authoritativeness | Is this a go-to source for the topic? | Reputation, citations and mentions from independent, reputable sources |
| Trust | Can I rely on this? | Clear authorship and ownership, accurate information, transparent policies, secure transactions |
Trust is the load-bearing signal. The guidelines say it plainly: "Trust is the most important member of the E-E-A-T family because untrustworthy pages have low E-E-A-T no matter how Experienced, Expert, or Authoritative they may seem." Their example is a financial scam run by a highly experienced, expert scammer. One sign you're hiding something cancels out a hundred signs you know what you're talking about.
E-E-A-T is a rater concept, not a single score in the algorithm. Google's systems use many signals that correlate with it, which is why working on it still pays.
YMYL: when the stakes get serious#
Your Money or Your Life (YMYL) topics are those that could significantly affect people's health, financial stability, safety, or the welfare of society. The September 2025 edition sorts them into four types of harm:
- YMYL Health or Safety: mental, physical and emotional health, and any form of safety, including online safety
- YMYL Financial Security: anything that could damage a person's ability to support themselves and their family
- YMYL Government, Civics & Society (renamed in September 2025): issues of public interest, trust in public institutions, and election and voting information
- YMYL Other: topics that could hurt people or the well-being of society
For YMYL topics the E-E-A-T bar is much higher, and rightly so. A blog post getting the dosage wrong for a common medication is a safety issue, not just a content quality issue.
What earns a "Highest" quality rating#
Raters score page quality on a scale from Lowest to Highest (Lowest, Low, Medium, High, Highest, with in-between ratings). Reaching the top tier requires:
- A clear, genuinely useful purpose for the visitor
- Very high E-E-A-T for the topic
- Enough effort, originality and depth to actually satisfy the query: not padded fluff, not three paragraphs that skip the hard parts
- A positive reputation among independent sources
- Accurate, current information
What gets you to the bottom? Deception, harmful content, spam (scaled content abuse, expired domain abuse, site reputation abuse), main content that is copied or AI-generated with little to no effort or added value, or simply: this page is obviously here to rank, not to help.
Needs Met: the second axis#
Page quality is only half the evaluation. Raters also score Needs Met: whether the result actually satisfies the query. A beautifully written, expertly researched article that doesn't address what the user searched for still gets a poor Needs Met score. The January 2025 edition expanded guidance on minor interpretations of ambiguous queries.
High Page Quality + high Needs Met = the combination that creates durable rankings.
AI-generated content: Google's current rules#
Google does not ban AI-assisted content. Its position, in "Google Search's guidance on using generative AI content" (first published May 2025, updated 1 October 2026):
- Using generative AI "to generate many pages without adding value for users" may violate the scaled content abuse spam policy. The page points to rater guideline sections 4.6.5 (scaled content abuse) and 4.6.6 (little effort, originality or added value).
- "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." The October 2026 update extends this explicitly to
<title>elements, meta descriptions, structured data and image alt text, because those appear in Search results too. - Give readers context on how content was made where it's useful. For ecommerce, Merchant Center requires AI-generated images to carry IPTC
DigitalSourceTypeTrainedAlgorithmicMediametadata and AI-generated titles and descriptions to be labelled.
The October 2026 update did not announce any new penalty or ranking signal; it wrote down a review step. Workflows for doing this well are in AI-Assisted Content Without Penalties.
Making E-E-A-T tangible#
Build real author pages#
Not just a name in a byline: a page with credentials, photo, links to their published work elsewhere, and a ProfilePage / Person JSON-LD block. Make it easy for raters and algorithms to verify that a real, qualified human wrote this.
Source everything#
Original research, quotes from recognised experts, links to the primary data your claims are based on. "Many experts believe..." is the reddest flag in content writing.
Show your work with experience#
Add genuine first-person perspective wherever it's credible: "I tested six tools and here's what I actually found." The Experience dimension matters more now that AI can produce fluent generic text: it can't fake having done the thing.
Keep it fresh#
Outdated information is a trust failure, especially for YMYL content. Surface "last reviewed" dates. Update statistics when the landscape changes. A 2019 tax guide still ranking in 2026 is a reputation risk. Content Refresh & Decay covers the process.
Maintain transparency pages#
A clear About page. Clear editorial and accuracy policies. Obvious disclosure when content is sponsored. These aren't SEO tricks; they're the basic signals of a trustworthy publication, and raters are trained to look for them.
Don't rent out your reputation#
Hosting third-party content mainly to exploit your domain's ranking signals breaks Google's site reputation policy (renamed from "site reputation abuse" in August 2026). Google now weighs whether partner content is presented consistently, matches your quality, has clear authorship, and isn't duplicated across other sites. Since 30 August 2026 the resulting manual actions apply outside the EEA only; inside the EEA, such sections may instead be ranked separately from the rest of the site. See Algorithm Updates and Penalty Recovery.
Next: Algorithm Updates, how Google applies quality signals at search engine scale, and what happens when an update reshuffles the deck.