LLM Best Practices

E-E-A-T: Experience, Expertise, Authoritativeness, Trustworthiness

3 min read

View markdown source

Overview

E-E-A-T is Google’s quality framework: Experience, Expertise, Authoritativeness, and Trustworthiness. It is not a ranking factor with a number attached; it is the rubric human raters use to grade search results, which then trains the ranking systems. Pages that demonstrate all four pillars rank better, especially on Your-Money-Your-Life topics (health, finance, legal). The four signals are independent; ship all four on every editorial page.

Show experience by saying who used the thing

The first E (added in late 2022) asks whether the author has hands-on experience with the topic, not just book knowledge.

  • Name the projects, versions, and dates the author worked with the tool. “Ran Postgres 16 in production at X from 2023 to 2025” beats “Postgres expert.”
  • Include first-person observations, screenshots from real dashboards, and benchmarks the author ran themselves.
  • Link to the author’s GitHub, conference talks, or published code that demonstrates the experience.

Experience is the cheapest pillar to fake and the easiest for raters to verify. Faked experience is detected through citation network analysis; recovery takes months.

Show expertise with credentials, depth, and named authors

Expertise is “does this author know the subject well enough to be cited?” Show it explicitly.

  • Byline every editorial page with a real human name. No “Staff” or “Admin.”
  • Author bio on every page: one paragraph, three to five facts, and a link to a full author profile at /authors/<slug>.
  • Author profile page lists credentials, prior publications, social profiles, and contact.
  • For YMYL topics, surface formal credentials (MD, CPA, JD) in the bio and in Person JSON-LD.

Mark up authors with Person JSON-LD

JSON-LD is how Google connects bylines to entities in its knowledge graph.

<script type="application/ld+json">
{
  "@context": "https://schema.org",
  "@type": "Article",
  "headline": "Postgres in Production",
  "datePublished": "2026-05-14",
  "dateModified": "2026-05-14",
  "author": {
    "@type": "Person",
    "name": "Jane Doe",
    "url": "https://example.com/authors/jane-doe",
    "sameAs": [
      "https://github.com/janedoe",
      "https://linkedin.com/in/janedoe",
      "https://x.com/janedoe"
    ],
    "jobTitle": "Staff Engineer",
    "worksFor": { "@type": "Organization", "name": "Example Inc." }
  }
}
</script>

The sameAs array is the entity-resolution signal; populate it with three to five public profiles per author. See the broader schema catalog in schema-markup-deep.

Earn authoritativeness through inbound citations, not self-claims

Authority is what other sites say about this one. The only direct moves are the ones that produce inbound citations.

  • Publish original research, benchmarks, or data. Other sites cite quotable numbers.
  • Ship reference tables and definitions designed for excerpting. See content on cluster strategy.
  • Earn backlinks from sites in the same topical neighborhood. Ten links from relevant sites beat 100 from unrelated ones.
  • Track referring domains, not just total backlinks. A new referring domain is a new authority signal.

Earn trust with citations, freshness, and operational hygiene

Trust is the meta-pillar; the other three feed into it.

  • Cite primary sources inline, with publication date and a real URL. See how-to-cite.
  • Disclose conflicts of interest. Affiliate disclosures live above the fold, not in a footer.
  • Show a real contact method, a privacy policy, and a physical address for the publisher.
  • Refresh and re-date pages when the body materially changes. Stale dates erode trust faster than no date.

Surface freshness signals where Google can read them

Google checks four freshness signals; ship all four when content changes.

  • dateModified in Article JSON-LD.
  • Visible “Last updated” line on the page, near the title.
  • <lastmod> in the sitemap.xml entry for the URL.
  • Updated last_updated in the page’s frontmatter, which drives the other three.

Faking freshness (re-dating without editing the body) is detectable through content-diff fingerprinting and triggers a downweight. Update the date only when the content has changed in a way that would change a reader’s decision.

Graph View

Backlinks