E-E-A-T audit
Talk like ../../references/plain-language.md. Call it "trust signals" first; expand E-E-A-T once.
Input
A public URL and/or pasted draft. If the fetch fails, ask them to paste the text.
Web connector mode
Score from pasted text or a public URL in chat. Optional style gate: check_ai_content. Paid SERP/backlink extras only after quote_cost and a yes. No Citebase fee for the trust audit itself.
Steps
- Read the page or draft. Note byline, dates, first-person detail, sources, schema hints, about/contact links, disclosures.
- Score each dimension 1–10 using
../../references/eeat-checklist.md. Quote evidence; do not invent credentials. - Apply vetoes (fake reviews, undisclosed paid placement, YMYL without qualified review, scaled unreviewed AI filler).
- Optionally run
python3 "${CLAUDE_PLUGIN_ROOT}/scripts/ai_content_lint.py" --jsonon the draft when text is available (web: callcheck_ai_content). - Deliver: per-dimension scores + top 5 fixes + verdict ship / fix then ship / rewrite. End with Done / Cost so far / Next.
Rules
- White-hat only. Never suggest fake authors, bought reviews or undisclosed affiliates.
- Local; no Citebase fee and no Apify spend unless the user separately asks for live SERP/backlink data.
- For AI citation goals after trust fixes, hand off to
ai-search-visibilityorcitation-loop.