Content Quality, E-E-A-T & YMYL

This chapter distills Google's Search Quality Rater Guidelines. Remember: ratings never directly move rankings — but they are the clearest public description of what the algorithms are trained to reward.

The core framework

Term	Definition
MC	Main Content — everything that directly helps the page achieve its purpose (text, media, features, and the page title).
SC	Supplementary Content — aids UX but doesn't achieve purpose (nav links).
Ads	Monetization (ads, affiliate, sponsored). Presence/absence is not itself a quality reason, but the site is responsible for overall ad quality.
E-E-A-T	Experience, Expertise, Authoritativeness, Trust (Trust is the center).
YMYL	Your Money or Your Life — topics that can significantly harm health, financial stability, safety, or societal welfare.

First step of any quality evaluation = determine the page's purpose. Pages should be created to help people; pages made only to benefit the owner with little attempt to help → Lowest. No page type is inherently higher quality (a humor page is not below an encyclopedia page) — purpose, not type, sets the ceiling.

MC quality — the four levers

Dimension	Guidance
Effort	How much a human actively worked. Auto-generating thousands of pages with no oversight = no effort. Forum/social discussion depth = cumulative human effort.
Originality	Unique content; if similar exists, is this the original source?
Talent / Skill	Enough to satisfy the purpose.
Accuracy	For informational/YMYL: factually accurate, consistent with well-established expert consensus.

High and low quality exist at all lengths and formats. Standards are purpose-relative (medical → consensus; how-to → skill; art → talent/originality).

E-E-A-T

• Trust is the center and most important. An untrustworthy page has low E-E-A-T no matter how expert it looks (a skilled scammer is still low E-E-A-T). Trust = accurate, honest, safe, reliable.
• Experience = first-hand/life experience. Expertise = necessary knowledge/skill (can be informal for niche fields). Authoritativeness = recognized go-to source (creator + content + site).
• Trust needs vary: stores need secure payment + reliable service; YMYL info must be accurate; casual entertainment needs little.
• Conflict of interest lowers Trust (a manufacturer "reviewing" its own product; paid-influencer reviews).
• Assess E-E-A-T from (1) what the site says about itself, (2) independent sources, (3) what's on the page.
• ⚠️ E-E-A-T is NOT itself a direct ranking factor — Google's systems use signals that correlate with it. Experience was added to E-A-T in Dec 2022.

Reputation research

Be skeptical of self-claims; find independent reviews, expert references, news, Wikipedia, forums. Research reputation for the specific topic. YMYL reputation should come from experts/professional societies. A large number of detailed credible positive reviews = positive; a few negatives are normal for any large business; read review content, not just stars. Absence of reputation is neither positive nor negative for ordinary/small entities. Search operators: [brand -site:brand.com], [brand reviews -site:brand.com], [brand site:en.wikipedia.org].

YMYL

Topics that could significantly impact health/safety, financial security, government/civics & society, or general societal welfare. Two harm mechanisms qualify a topic: (1) the topic is itself harmful (self-harm, violent extremism), or (2) it could cause harm if inaccurate (heart-attack symptoms, how to invest, who can vote). YMYL is a spectrum. Diagnostic: would a careful person seek experts to prevent harm? → likely YMYL. Clear YMYL gets the most scrutiny and highest PQ standards.

Personal life experience on a YMYL topic (coping with cancer, first-time voting) can be high E-E-A-T if trustworthy/safe/consistent with expert consensus — but medications, treatment options, tax-form instructions, how-to-invest require experts only.

The Page Quality scale (Lowest → Highest)

Slider: N/A · Lowest · Lowest+ · Low · Low+ · Medium · Medium+ · High · High+ · Highest. Torn between two → use the lower; among three → the middle. Medium is very common.

Three-step process: (1) assess true purpose — deceptive/harmful purpose → Lowest; (2) assess harm potential — harmful/untrustworthy/spammy → Lowest (a harm screen applied before other considerations, even on govt/academic/reputable sites); (3) otherwise rate by how well it achieves its purpose.

Lowest (any one suffices): hacked/gibberish MC; MC with little effort + no originality + no added value; extremely misleading/exaggerated title; MC obstructed by ads/interstitials benefiting the owner; complete lack of ownership info on trust-requiring/YMYL pages; very negative reputation; untrustworthy page whose lack of E-E-A-T causes purpose failure. Rate Lowest even if you can't "prove" harm.

Low (any one suffices): inadequate effort/originality/talent to satisfy purpose; mild inaccuracies on informational pages; slightly misleading/exaggerated title; distracting ads/SC that interrupt MC; "filler" that pushes helpful MC down (recipe buried under filler); inadequate site/creator info; inadequate E-E-A-T for the purpose — positive reputation or site type cannot overcome inadequate E-E-A-T; low-value repackaging ("best" lists from existing reviews).

Medium: beneficial purpose achieved, adequate MC/E-E-A-T, title summarizes, ads don't block MC. Either "nothing wrong but nothing special" or "mixed High + mild Low signals." A new unanswered forum question by a real person → generally Medium.

High (≥1): high effort/originality/talent/skill; positive reputation; high E-E-A-T. Signals: well-edited/curated, original photos/footage, first-hand experience, skilled how-to. Small sites with no reputation info can be High.

Highest (≥1): very high effort/originality/talent; very positive reputation; very high E-E-A-T (uniquely authoritative go-to source). Examples: award-winning investigative journalism; official govt sources (Naval Observatory clock, NPS, IRS tax forms, annualcreditreport.com); kernel.org; top-hospital medical pages; bank login; a blogger's extensively-tested recipe (first-hand experience).

Lessons from the visual examples

• PQ judges the page in isolation (trust, effort, E-E-A-T, reputation); NM judges fit to a specific query's intent. A page can be high PQ yet FailsM for a mismatched query, and vice versa.
• Lowest figures are driven by untrustworthy/deceptive/spammy signals: fake ".gov" credit-report page with gibberish MC; "Buy Xanax" keyword-stuffed spam; auto-generated error-code pages with visible blank template fields ("the reason should be ."); expired-domain abuse (rural-school nonprofit → lottery, caught via Wayback); site-reputation abuse (advertiser "article" under a news banner); harmfully inaccurate YMYL medical Q&A.
• Low = unsatisfying/low-effort, low E-E-A-T but not deceptive: garbled health article with no expertise; generic "how to adopt from Iraq" with nothing Iraq-specific; recipe buried under filler/ads (obstruction is the demotion lever — a prominent recipe would rate higher); financial article with no expertise.
• Medium = competent but unremarkable: a bare recipe with no original photos/reviews caps at Medium even from a credentialed author (authorship alone doesn't lift a thin page); a bare custom-404 with only a home link.
• High = original MC + reputation + demonstrated E-E-A-T: professional reporting; official govt page; expert-bylined blog post; reputable store with reviews + service info; "An Engineer's Guide to Cats" (firsthand experience + talent, no formal credentials needed for entertainment); go-to niche forum.
• Highest = comprehensive/uniquely authoritative/award-worthy: investigative journalism on YMYL; uniquely authoritative primary sources; authoritative medical tools; a manufacturer selling its own unique product with 600 genuine reviews; even amateur work that is uniquely original + high-effort + award-recognized ("Henri, Paw de Deux"). Amateur ≠ low.

Needs Met (how a result fits a query)

Slider: N/A · FailsM · SM · MM · HM · FullyM.

Rating	Meaning
Fully Meets	Special — only for clear single-result intent where one specific site/page/fact satisfies all/almost-all users ([amazon]→amazon.com; [what country is mount fuji in]→Japan). Verify facts. Most queries can't have FullyM.
Highly Meets	Very helpful for a dominant/common/reasonable-minor interpretation. Often the highest possible rating. Accuracy required; YMYL HM must reflect expert consensus.
Moderately Meets	Helpful but fewer valuable attributes; nothing wrong, average-to-good.
Slightly Meets	Less helpful, or helpful for an unlikely minor interpretation; may be outdated/indirect.
Fails to Meet	Off-topic; wrong location/intent; incorrect/very outdated; harmful/porn/unwanted.

Key relationships: NM depends on query+result; PQ does not depend on the query. "Useless is useless" — a useless result is FailsM even with high PQ (exception: clear website-intent queries → target site = FullyM). PQ can cap NM on YMYL (a no-expertise health page → MM at best; Mayo Clinic → very helpful). HM is inappropriate for untrustworthy/outdated/inaccurate pages. Misleading/exaggerated titles → SM or lower. Give HM broadly across diverse interpretations, formats, and depths for exploratory queries.

Freshness: for breaking news/recurring events/current-info/product queries, assume users want the latest — stale content = low NM (even one-day-old traffic). Freshness is less of a PQ concern (archival content can be High PQ; unmaintained/abandoned = low PQ). Check content history via the Wayback Machine.

Generative AI content stance

• No blanket ban. AI/automation is fine if the output is helpful, original, people-first. Evaluation is about value and quality, not production method. High-effort original work made with AI can be High/Highest.
• It becomes a violation when used to scale content to manipulate rankings (scaled content abuse) → Lowest regardless of method.
• AI tells that signal Lowest/deceptive: leftover "As an AI language model," "my knowledge cutoff is September 2021," abrupt mid-sentence cutoffs, odd errors no human would make ("praying mollusks"), templated odd Q&A, fake AI-generated author profiles/bios (deceptive → Lowest), terms-of-use disclaimers admitting "some articles may be AI-generated… may have errors" (untrustworthy → Lowest, especially YMYL).
• Disclose AI/automation when a reader would reasonably wonder "how was this made?" Make Who / How / Why clear. (Bing aligned to this Google stance Feb 2026 — Chapter 11.)

Related on Crawlinx

• E-E-A-T SEO: Experience, Expertise, Authority, Trust
• Technical SEO Audit Checklist: 40+ Checks for 2026
• content.thin
• content.low_ratio
• eeat.no_author
• eeat.no_date

Sources

1. Google Search Quality Rater Guidelines (PDF)
2. Google Search Central — Creating helpful, reliable, people-first content
3. Google Search Central Blog — Experience added to E-A-T
4. Google Search Central — Guidance on AI-generated content