EEAT in 2026: How Google Quality Raters Score Your Site

First published: 14 May 2026 · Last updated: 14 May 2026

T

Trust

The most important member of the family. Accuracy, transparency, safety. Without Trust, the other three do not save the page.

E

Experience

First-hand involvement with the topic. Added Dec 2022, elevated Sep 2025.

E

Expertise

Topic knowledge appropriate to the page's purpose. YMYL pages need formal expertise.

A

Authoritativeness

Reputation of the creator and the site as a go-to source for the topic.

E-E-A-T has become one of the most misused acronyms in SEO. Half the agency content treats it as a direct ranking factor (it is not). The other half treats it as marketing fluff (it is not that either). The honest framing is that E-E-A-T is the evaluation framework Google's roughly 16,000 human Search Quality Raters apply when scoring page samples, and those scores are used to train and validate the ranking algorithms. It is one step removed from ranking, but it shapes the signals the algorithm learns to reward. This article is the practitioner version. We work through what changed in the September 2025 Quality Rater Guidelines (QRG) update, how raters actually score sites in practice, the YMYL distinction that doubles the scoring stakes, and the practical signals to deploy. If you want the surface-level version, our older Google E-A-T explainer covers the basics. This one is for SEO leads who need to brief content teams on exactly what to build.

What the Quality Rater Guidelines Actually Are

The Quality Rater Guidelines is a 180-page document Google publishes (most recently updated 11 September 2025) that instructs roughly 16,000 contracted human raters on how to evaluate the quality of search results. Raters do not influence rankings directly. They score sample queries and pages, and Google's machine learning systems use those scores as ground truth to train and evaluate the ranking algorithms. The mechanism matters because it explains the lag. A change to the QRG (like the Sep 2025 update) does not move rankings overnight. It changes the training signal that the next round of algorithm updates is calibrated against. Expect to see the rated changes show up in rankings 3 to 9 months after a QRG update, often bundled into a Core Update or Helpful Content System refresh.

What Changed in the September 2025 QRG Update

The September 2025 update was the most significant E-E-A-T revision since Experience was added in December 2022. Four shifts matter for SEO professionals: 1. Experience evidence is now scored more strictly. Pages claiming first-hand experience without supporting signals (original photos, named author with verifiable involvement, specific dated events) are downgraded more aggressively. Generic "based on our research" language no longer counts as Experience evidence on its own. 2. AI-generated content is explicitly addressed. The QRG now contains guidance for raters on identifying scaled, automated content created with minimal human oversight. Such content can still be rated High or Highest if it provides genuine value, but the bar for Trust evidence rises sharply when AI authorship is detected or suspected. 3. Trust evaluation tightened for YMYL. Your Money or Your Life topics (health, finance, legal, civic) now require explicit credential signals. A health article from an unnamed author or an author with no verifiable medical background is rated Lowest by default, regardless of content quality. 4. Site reputation is weighted more heavily than individual page quality. A high-quality page on a low-reputation site is now capped lower than the same page on a high-reputation site, making domain-level authority work (Tactic 7 in our GEO playbook) more important.

How Raters Actually Score a Page

The QRG defines two principal scoring scales raters apply per page sample. Understanding these scales is the difference between guessing at E-E-A-T and engineering it.

Page Quality (PQ) rating

Lowest : Untrustworthy, harmful, deceptive, or zero E-E-A-T

Low : Lacking E-E-A-T, poor reputation, or thin content

Medium : Adequate E-E-A-T, achieves purpose without distinction

High : Strong E-E-A-T, achieves purpose well, positive reputation

Highest : Exceptional E-E-A-T, very high quality MC, strong reputation

Needs Met (NM) rating

Fails to Meet : Result is unhelpful or unrelated to query intent

Slightly Meets : Helpful for some users, low effort or quality

Moderately Meets : Helpful for many users, average satisfaction

Highly Meets : Very helpful for most users, high satisfaction

Fully Meets : Definitive answer for query intent, no further search needed

PQ is the page-quality side, evaluated in isolation from any specific query. NM is the query-relevance side, evaluated against a specific user intent. A page can be Highest PQ but Slightly Meets NM if it is not actually relevant to the query the rater is testing. A page can be High NM but Low PQ if it answers the query but is from a low-quality source. Both scores feed back into algorithm training. For SEO purposes, PQ is the structural work (E-E-A-T evidence, site reputation, content quality, author transparency) that travels across all queries the page might rank for. NM is the per-query intent matching work (does this page actually answer the query the user typed). The two compound.

The YMYL Multiplier

YMYL (Your Money or Your Life) topics carry double the scoring stakes. The QRG defines YMYL as topics that could significantly impact a person's health, financial stability, safety, or societal welfare. Examples include medical advice, financial guidance, legal information, news on civic matters, and child-safety content. For YMYL pages, the QRG instructs raters to apply higher E-E-A-T standards. Specifically:

• Author credentials must be visible and verifiable. A medical article by "Dr Tan" with no biography is rated Low by default. The same article with a linked Person schema, professional registry link, and visible credentials can be rated High.
• Trust signals are compulsory, not optional. Citations to authoritative sources, named publication date, named editorial review, and clear corrections policy.
• Site reputation matters more. A YMYL page on a known reputable site (named medical publication, government body) is rated higher than the same page on an unknown blog, even with identical content.

A practical Singapore-specific note: medical and financial sites operating under MOH or MAS regulatory oversight should make that compliance visible on every YMYL page (license number, regulator badge, named compliance officer). This converts an external trust signal into a page-level E-E-A-T signal that raters can verify.

The Practical Signal Stack

Translating QRG scoring into deployable signals on your site is the hard part. Here is the prioritised checklist, grouped by the four pillars.

Experience signals

• Original photography of the subject (product unboxing, location visit, event attendance)
• First-person prose with specific dates, places, named encounters
• Author bio that names the relevant first-hand experience ("ran 12 SEO audits for SG B2B sites in 2025")
• Original screenshots, dated process logs, behind-the-scenes content
• Video evidence of the author engaging with the subject where applicable

Expertise signals

• Author credentials in bio (degrees, certifications, professional memberships)
• Person schema with jobTitle, alumniOf, knowsAbout, sameAs to LinkedIn or ORCID
• Author has published multiple articles on the same topic on this domain
• Citations of and engagement with primary sources, not just other blogs
• For YMYL, professional registry verification visible on the author page

Authoritativeness signals

• External brand mentions in trusted publications (without requiring backlink)
• Author byline on trusted external publications, linked from author bio
• Wikipedia or Wikidata entry for the author or organisation where eligible
• Conference speaking, podcast guesting, named research datasets
• Industry awards or recognised certifications listed and verifiable

Trust signals

• Visible business name, address, registration number on Contact page
• HTTPS site-wide, valid SSL, no mixed content
• Visible publication date and last-updated date on every article
• Editorial policy and corrections policy linked from footer
• Named editorial reviewer for YMYL content
• Privacy policy, terms of service, cookie policy current and accessible
• For ecommerce: returns policy, customer service contact, real reviews with verification

The Author Page: The Single Highest-Leverage E-E-A-T Asset

If you do one E-E-A-T deployment, build proper author pages. The author page is where Experience, Expertise, and Authoritativeness signals concentrate, and the QRG explicitly directs raters to look up the author when evaluating a page. A QRG-compliant author page contains, at minimum:

• Full name, professional headshot, current role.
• Biography that names specific first-hand experience (not generic "passionate about marketing" filler). The bio should state what the author has actually done, with specifics.
• Credentials section listing degrees, certifications, professional memberships, with verification links where possible.
• External validation links: LinkedIn (mandatory), Wikipedia or Wikidata where eligible, ORCID for academics, X, GitHub, professional registry pages, conference speaker pages.
• Published article archive showing the author's body of work on this site, demonstrating consistent topical coverage.
• Contact mechanism so raters can verify the author exists and is reachable.
• Person schema with jobTitle, worksFor, alumniOf, sameAs array, knowsAbout array, image, url.

Once author pages are in place, every article must carry a visible byline that links to the author page, and the Article schema must reference the author entity by URL. This closes the loop from article to author to verifiable expertise. For our own author setup at BestSEO, we treat the author page as a first-class SEO asset, not an afterthought, and we measure its outbound clicks alongside traditional ranking metrics.

Site-Level Reputation: What Raters Look For

The QRG instructs raters to assess site-level reputation by looking beyond the site itself. The standard rater process is to search "[brand name] reviews", "[brand name] reputation", and similar queries, then read what independent sources say. This means your E-E-A-T strategy cannot be entirely on-site. The off-site reputation footprint must be coherent. Specifically, raters look for:

• Independent reviews on Google Business Profile, Trustpilot, G2, Capterra, industry-specific review sites. Review velocity and recency matter.
• News mentions on credible publications. Even neutral coverage (you were quoted in a Reuters article) counts as authority signal.
• Forum and community discussion on Reddit, Quora, industry forums. Raters read these. Negative discussion is not automatically disqualifying, but consistent quality complaints are.
• Wikipedia and Wikidata presence where eligible. These act as independent authority anchors for the brand entity.
• BBB rating (US) or equivalent (ACRA registration in SG) for legitimacy verification.

A practical action item: search your own brand name in incognito and read what comes up on the first two pages. That is approximately what a quality rater sees. If those pages do not paint a credible, professional picture, your site-level reputation is bottlenecking your E-E-A-T regardless of how well your on-site signals are deployed.

E-E-A-T and AI-Generated Content

The September 2025 QRG update added explicit guidance on automated and AI-generated content. The key principle: AI-generated content is not automatically penalised, but it raises the Trust bar. The QRG distinguishes between:

• AI-assisted content with substantive human editing, review, and verification. Treated equivalently to human-written content. E-E-A-T evaluated normally.
• Scaled, automated content created with minimal human oversight. Rated Lowest by default unless it provides genuinely useful, original information that adds clear value beyond what existing sources offer.
• Deceptive AI content that misrepresents itself as human-authored or fabricates author credentials. Always rated Lowest.

For SEO teams using AI in production, the operational implication is clear: every AI-assisted article needs a named human editor, evidence of human review (specific edits, not boilerplate disclaimers), and Trust signals (sources, dates, editorial policy) that match or exceed human-authored content. Disclosing the use of AI assistance is permitted but not required, provided the content meets the Trust bar. This dovetails with the Information Gain principle we covered in the GEO playbook (Tactic 3): regurgitated AI content fails both the GEO citation test and the QRG Trust test. Original frameworks, original data, and named expert quotes pass both.

How E-E-A-T Interacts with the Helpful Content System

The Helpful Content System (HCS), now folded into Google's core ranking system as of the March 2024 Core Update, is the algorithmic counterpart to the QRG's Page Quality scoring. Both are looking at the same underlying signals from different angles: the QRG is human-evaluated, the HCS is algorithmic, but they triangulate on the same target.

Month 0

QRG update published

Google releases revised Quality Rater Guidelines. Raters retrained on new criteria within weeks.

Month 1-3

New training data accumulates

Raters score thousands of new page samples under updated criteria. ML systems ingest the new ground truth.

Month 3-6

First Core Update calibrated on new signal

The next Core Update typically reflects partial calibration. Early movers (sites already aligned) see lift.

Month 6-12

Full propagation

Subsequent Core Updates compound the effect. Sites that ignored the QRG shift see sustained suppression.

Practical consequences:

• A site with weak E-E-A-T but strong topical content can still get HCS-suppressed if the algorithm detects scaled-content patterns or thin author signals.
• A site with strong E-E-A-T but weak topical fit will not be artificially boosted by E-E-A-T signals alone. E-E-A-T is a multiplier on relevance, not a substitute for it.
• The HCS is site-wide, not per-page. A few low-E-E-A-T pages in a YMYL category can pull the entire domain's HCS scoring down.

For sites recovering from a HCS-related visibility drop, audit at the site level first (author transparency, Trust signals, editorial policy, AI content review process), then per-page (Experience evidence, primary sources, named expertise). Single-page edits rarely move HCS scoring; site-level structural changes do.

E-E-A-T for AI Search (GEO Implications)

E-E-A-T is no longer just a Google Quality Rater concern. AI engines (ChatGPT Search, Perplexity, Claude, Google AI Overviews) apply analogous trust scoring during their retrieval and synthesis pipelines. Pages that score well on E-E-A-T signals are disproportionately cited in AI-generated answers, even controlling for content relevance. The mechanism is most explicit in Google AI Overviews, which inherits the underlying ranking signals (and therefore E-E-A-T-derived weights). It is also visible in Claude, which we documented in our multi-engine ranking playbook as the engine most biased toward authoritative publications and credentialed sources. The practical synthesis: an E-E-A-T programme in 2026 is not separate from a GEO programme. They are the same programme viewed from two angles. The author page work, the Person schema, the Trust signals, the site reputation work, the editorial policy, all of it, compounds in both classical SERPs and AI engine citations.

Frequently Asked Questions

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

No. E-E-A-T is the framework Google's human Search Quality Raters use to evaluate page quality. Their ratings train and calibrate the ranking algorithms but do not directly influence individual page rankings. The practical effect is that strong E-E-A-T signals correlate with ranking strength because the algorithms are trained to reward the same signals raters reward, but no single E-E-A-T element can be pointed to as a ranking factor in the way that, say, page speed can.

What is the difference between Experience and Expertise?

Experience is first-hand involvement with the subject (the author has used the product, visited the location, lived through the event). Expertise is topical knowledge appropriate to the page's purpose (the author has studied or practised in the field). A travel blog post about a specific hotel needs Experience (you stayed there) more than Expertise (formal hospitality qualifications). A medical advice page needs Expertise (medical credentials) more than Experience (you happen to have had the condition). Most pages benefit from both.

How important is the Trust pillar relative to the others?

Trust is explicitly identified in the Quality Rater Guidelines as the most important pillar. The other three (Experience, Expertise, Authoritativeness) function as evidence supporting Trust. A page can have strong Experience, Expertise, and Authoritativeness, but if Trust signals are missing or compromised (no contact details, no editorial policy, deceptive elements, weak Trust hygiene), the page is rated Low or Lowest regardless. Build Trust first, layer the other three on top.

How does YMYL change E-E-A-T scoring?

YMYL (Your Money or Your Life) topics raise the E-E-A-T bar substantially. For YMYL pages, raters require explicit author credentials, visible publication and review dates, citations to authoritative primary sources, and high site-level reputation. A YMYL page from an anonymous or under-credentialed author is rated Low by default, even if the content is accurate. Non-YMYL pages have more latitude: a recipe blog with no medical credentials is fine, but a recipe blog page about diabetes-friendly meals tips into YMYL territory and the rules tighten.

How long does E-E-A-T work take to show up in rankings?

Three to nine months is typical. The QRG drives algorithm training, which manifests in rankings via Core Updates, which Google releases roughly every two to four months. A deployment shipped in May 2026 usually shows visibility movement at the next Core Update, with the full effect compounding over the subsequent two updates. Plan E-E-A-T programmes on a 6 to 12 month horizon, not weekly sprint cycles.

Should I disclose AI usage in my content?

Disclosure is permitted but not required, provided the content meets the Trust bar of human-authored content. The QRG's distinction is between AI-assisted content with substantive human editing (treated normally) and scaled automated content with minimal human oversight (rated Lowest). What matters operationally is that every AI-assisted article has a named human editor, evidence of substantive review, and Trust signals (sources, dates, editorial policy) that meet or exceed human-authored equivalents. Many publishers add a brief disclosure for transparency, but it does not in itself confer or remove E-E-A-T credit.

Related reading

• The GEO playbook: 9 optimisation tactics that move AI citations
• Google E-A-T: the original explainer (now superseded)
• Multi-engine ranking playbook: ChatGPT, Perplexity, Claude, AI Overviews