seo-eeat-ymyl-google

name	seo-eeat-ymyl-google
description	Apply when a site-wide audit according to Google Search Quality Rater Guidelines is needed with a focus on E-E-A-T, YMYL, Page Quality, and Needs Met for medical, financial/transactional, and standard content sites with a scorecard-first output.

SEO E-E-A-T YMYL Google

Using seos-cli in this skill

Adoption decision: seos-cli is optional and not used by default. This skill's core work is QRG judgment, hard-gate analysis, E-E-A-T scoring, and evidence-matrix synthesis; seos-cli does not perform that reasoning.

Use seos-cli only as supplemental evidence collection when project-native page text or SERP evidence would materially help the audit:

seos-cli config check
seos-cli scrape --urls https://example.com/page --format json --out outputs/eeat-page-scrape.json
seos-cli serp --keywords "brand query, product query" --format xlsx --out outputs/eeat-serp-evidence.xlsx

Do not use seos-cli when local HTML, WebFetch, a subagent scrape, reputation research, or scripts/extract_eeat_signals.py already provides the evidence needed by this skill. Do not claim seos-cli performs E-E-A-T scoring, YMYL classification, Page Quality, Needs Met, or final QRG verdicts.

Overview

QRG-first skill for site-wide quality, trust, and risk audit according to Google Search Quality Rater Guidelines 2025.

Core principle: hard-gates and evidence first, then Page Quality, then E-E-A-T scorecard. Needs Met is evaluated only if a query context is available.

When to apply

• A site audit or URL sample audit is needed for E-E-A-T, YMYL, Page Quality, trust, reputation, harm.
• Need to separate low-trust YMYL/transaction pages from standard informational pages.
• Need to generate a scorecard-first report, not a narrative without metrics.
• Need to quickly and repeatedly extract on-page signals before manual judgment.

Do not use as a crawl-heavy technical audit. For leak/proxy models, graph scoring, and massive crawling, use other skills.

Discovery scenarios

• Medical site: check YMYL-risks, trust-signals, author, and review process.
• Financial/transactional site: check safety, transaction transparency, and customer-service trust signals.
• Standard content site without obvious YMYL: check usefulness, content quality, and basic trust.

Expected input

• Domain or list of URLs.
• If available: query or query-cluster for Needs Met.
• If necessary: local HTML files for selective checking.

Workflow

1. Determine the site type and YMYL Tier: Non-YMYL | Possible YMYL | Clear YMYL.
2. Collect a sample of 5 URLs: homepage, money/service/product, informational, trust/about/contact/author, highest-risk YMYL or transaction.
3. For webfetch, scraping, external reputation search, and mass page reading, invoke a subagent first.
4. Pass hard-gates: harmful purpose, deceptive purpose/design, scaled low-value content, reputation abuse, hidden/blocked main content, unsafe sensitive-data requests, untrustworthy YMYL claims.
5. If necessary, run scripts/extract_eeat_signals.py to get a compact JSON for the pages.
6. Assign Page Quality: Lowest | Low | Medium | High | Highest.
7. Fill out the E-E-A-T scorecard: Experience, Expertise, Authoritativeness, Trustworthiness, Overall E-E-A-T.
8. If there is a query context, evaluate Needs Met separately.
9. Generate the final report according to references/output-contract.md. If desired, use scripts/render_scorecard.py only to prefill the template, not for final judgment.

Mandatory rules:

• Do not mix Page Quality and Needs Met.
• If query context is missing, set N/A for Needs Met.
• hard-gate findings can forcefully lower the result regardless of the formal score.
• Explicitly distinguish Experience and Expertise; in high-risk YMYL, do not substitute formal expertise with personal experience.
• scripts/render_scorecard.py can be used only for prefill and structure normalization, but not for automatic verdicting.

What to read

• references/qrg-page-quality.md: hard-gates and Page Quality logic.
• references/ymyl-thresholds.md: YMYL Tier, trust requirements, experience vs expertise.
• references/eeat-scorecard-model.md: rubric for E-E-A-T scoring.
• references/needs-met-and-intent.md: when and how to use Needs Met.
• references/qrg-metric-catalog.md: full metric register that must be included in the report.
• references/output-contract.md: mandatory sections, evidence format, and wording rules.

Lite Scripts

scripts/extract_eeat_signals.py

• Extracts on-page signals from URLs or local HTML.
• Returns a compact JSON for scorecard prep.

scripts/render_scorecard.py

• Renders the markdown report template and evidence matrix.
• Does not make final decisions instead of the agent.

Language and context rules

• Write the main text of the skill, reference files, templates, and report in English.
• Use English terms only as standard labels: E-E-A-T, YMYL, Page Quality, Needs Met, Lowest, Highest, Person schema.
• Do not create bilingual duplicates.
• Delegate long scraping/webfetch/reputation-heavy stages to a subagent; the main agent receives only a compressed summary, extracted signals, and findings.
• Keep decision-making, scorecard assembly, and final report synthesis local.