The author entity — why the author, not the article, is the lever for E-E-A-T.

The question every SEO team finds itself asking after the third core update of 2023 at the latest: why do we keep losing rankings on YMYL-adjacent topics despite technically clean execution, strong backlinks and topically relevant content? The answer is rarely the article; it is the ontology behind it. Google's rater guidelines do not judge quality on the single document alone but explicitly ask about the content creator and the website. An article without an attributable author with verifiable expertise offers little evidence for that, no matter how well the content is written.

Does Google evaluate E-E-A-T per author or per article?

Neither, strictly speaking: according to Google, E-E-A-T is not a specific ranking factor but an evaluation framework that looks at the author, the page and the website together. Google's systems use a mix of factors that can identify content with good E-E-A-T, and for YMYL topics such as health, finances or safety they give such content even more weight (Google Search Central). The Search Quality Rater Guidelines (version of September 11, 2025) define Experience and Expertise explicitly in terms of the content creator: does the creator have first-hand experience, and the necessary knowledge? Authoritativeness applies to the creator or the website, Trust to the page itself, and Trust is the most important member. In practice, an article benefits from an author whose experience and expertise can be verified outside the article. That is exactly what an author entity with byline, author page, Person schema and independent evidence provides. Google says the rater assessments themselves do not directly influence ranking.

The Experience dimension added in December 2022 makes the logic even more explicit: Experience describes whether content was produced with first-hand experience, such as actual use of a product. That is something you mostly judge at the author level. An article on laser eye surgery written by an operating ophthalmologist carries a different Experience weight than the same text written by a content writer without medical practice. It only becomes machine-traceable when the author entity is linked to the corresponding credentials, published expert articles and verifiable roles. In Article markup, Google recommends an author URL or sameAs that uniquely identifies the author (author markup best practices).

The seven layers of a strong author entity

Layer 1 — Stable @id URI. Every author receives a canonical Person schema instance at a stable URI — typically https://brand.com/#author-john-doe or as a dedicated author page /team/john-doe/. The @id is the anchor node every Article schema references through the author property. Without that stable reference, author mentions remain lexical strings without semantic linkage, a structural weakness we have seen more and more often as a limiting factor in our audits since 2023.

Layer 2 — Person schema with a complete property set. The schema block carries name, jobTitle, description, url, image (with ImageObject nested plus caption and license), sameAs array (7–10 curated, authoritative third-party profiles), knowsAbout (topic tags for subject-matter expertise), knowsLanguage, worksFor (as an Organization @id reference), alumniOf (also @id-linked) and hasCredential for formal qualifications. Collectively, these properties form the machine-readable biographical signature. The denser and more referenced the property set, the higher the entity-resolution confidence.

Layer 3 — sameAs cluster across authoritative platforms. sameAs is the single most important property. It links the internal @id to external, verifiable identities. For an SEO strategist, that typically means LinkedIn, Crunchbase, GitHub, YouTube channel, speaker-platform profiles (Notist, SpeakerHub), Wikipedia article (where it exists), Wikidata Q-ID. For academic authors, add ORCID, Google Scholar, ResearchGate, Semantic Scholar, Crossref author profiles, VIAF. For journalists, add Muck Rack, author profiles in trade media, ProLit profiles, Mastodon accounts. What matters is not volume but consistency: identical name, identical photo, identical role description, identical company reference, consistent biographical statements across every platform. A single deviation, such as an incorrect company assignment on one trade-media profile, can noticeably weaken entity resolution.

Layer 4 — Wikidata item (where notability allows). For authors with external notability, the Wikidata item with a Q-ID is the structurally most important node. Google names hundreds of web sources, including licensed data, as the origin of Knowledge Graph information (Google blog, 2020); Wikidata is one of the open sources among them and is freely reusable under CC0 (Wikidata: Licensing). How much Wikidata flows into the training data of ChatGPT, Claude or Gemini is not disclosed by the providers. The item should contain at least 15 maintained properties with external references. For content writers without external notability, work on sameAs and corroboration matters more — a forced Wikidata item without serious, publicly available references can be removed through a deletion request under the Wikidata notability policy and leaves a deletion history.

Layer 5 — Credential provenance with EducationalOccupationalCredential. The decisive lever for YMYL topics. Schema.org credential objects link to the issuing institution, describe the competency type (degree, certification, license) and reference verification URLs. Example: hasCredential on a Person pointing to an EducationalOccupationalCredential of type "MD" from a specific medical school with a recognizedBy reference to a medical board. This chain makes qualifications machine-verifiable, which matters especially for YMYL topics, where Google gives strong E-E-A-T more weight.

Layer 6 — Corroboration through trade-media mentions. External, independent confirmation of the author's role — an author cited as a source in trade media, appearing as a speaker at industry conferences, contributing as a guest author in authoritative publications accumulates corroboration signals. Every external confirmation should link back to the author's own @id — ideally through Schema.org author linkage on the third-party publication, practically often through a consistent name fingerprint plus sameAs coherence.

Layer 7 — Publishing principles and organizational context. Schema.org publishingPrinciples links to a page describing the editorial standards of the publishing organization — fact-checking processes, correction policies, conflict-of-interest rules. For YMYL publishers, we consider it strongly advisable. The author entity is thereby placed in the context of an organization entity with explicit quality standards — a structural E-E-A-T signal that individual articles cannot produce.

Building the author entity: a 120-day protocol

Days 1–15 — audit and name-collision check. Collect every existing mention of the author, identify name collisions with other namesakes (Google name search, LinkedIn, ResearchGate). For relevant collisions, define a disambiguation strategy: introduce a middle name, carry an academic title consistently, anchor a role descriptor as a suffix. Baseline documentation of every existing third-party profile, including inconsistent data.

Days 16–45 — clean up and extend the sameAs cluster. Establish consistency across every existing profile: identical, high-resolution photo, identical name, identical role description, identical company assignment. Add missing authoritative profiles (ORCID for publishing authors, Crossref author profile for scholarly publications, Muck Rack for journalists). Target: 7–10 curated, consistent third-party profiles as sameAs candidates. Quality over quantity.

Days 46–75 — schema implementation and @id graph. Implement Person schema with a complete property set on the own domain, build a canonical author page with @id anchor, reference Article schemas across every publication to the author @id. Update the sitemap for the author page and strengthen internal linking from every article to the author page.

Days 76–90 — Wikidata item (if notability is given). Preparation: collect 10–15 independent references for the properties to be documented. Create the item with at least 15 properties (P31, P106, P108, P69, P1416, external identifiers). Invite third-party editors to review. Activate a watchlist for drift control.

Days 91–120 — corroboration initiative. Targeted placement of author mentions across three to five authoritative trade-media outlets with consistent biographical facts. Document speaking appearances at industry conferences and feed them into the credential chain. Set up and link the organization's publishing principles page.

The author entity in the LLM era

The relevance of the author entity escalates with LLM search. When a user asks ChatGPT, Claude or Perplexity, "Who is the leading expert on topic X?", the models draw on entity-graph signals, sameAs coherence and corroboration density. Authors with a strong structured presence get cited consistently; authors without remain either invisible or hallucinated. This is not an abstract problem: in advisory practice, we regularly see cases where LLMs misattribute authors — assigning one author the articles of another, or misrepresenting their role. The cause is almost always an inconsistent or missing author entity structure.

For content brands, that has strategic consequences. Instead of treating author bylines as a cosmetic feature, authors must be maintained as long-term entity assets. A brand with three strong author entities in its field accumulates E-E-A-T signals that isolated anonymous articles will never reach. The investment in author visibility pays off three times over: Google rankings, AIO citations and LLM citations — all draw from the same entity substrate.

Common mistakes when building the author entity

Six patterns that recur stubbornly in advisory practice and act as structural limits.

First: ghostwriter collisions. When the byline author is not the actual author, the Experience signal breaks. Google and LLMs detect inconsistencies between declared credentials and the actual publication pattern. Transparency (ghostwriter disclosures, co-author structure) beats fiction.

Second: author page without @id anchor. A bio page at /team/john-doe/ without an explicit schema @id anchor works for humans, not for machines. The @id anchor with a #person fragment is the actual node Article schemas must reference.

Third: weak sameAs cluster. Three third-party profiles are not enough, but curation beats volume: 7–10 genuinely authoritative, consistent identity profiles beat any long list of weak directory entries. And: inconsistent third-party profiles (outdated role on LinkedIn) hurt more than missing additional profiles.

Fourth: credential chain without references. EducationalOccupationalCredential objects without a recognizedBy or verifiable URL reference are worthless signals. Every credential needs an anchor back to the issuing institution.

Fifth: Wikidata item without maintenance. One-time Wikidata items left unmaintained for years drift through third-party edits and become a liability rather than an asset through outdated facts. Quarterly maintenance is mandatory.

Sixth: missing organizational context. An author entity without a worksFor reference to an Organization entity is inconsistent. Google expects an anchoring to a publisher context with its own E-E-A-T signals.

Measurement: quantifying the strength of an author entity

Four indicators with concrete tests; these are practical proxies from advisory work, not official Google metrics. (a) Google Knowledge Graph Search API: the author is returned as a node with their own @id and score — if not, entity resolution is missing. (b) LLM resolution test: a prompt matrix of 20 biographical questions across ChatGPT, Claude and Gemini — are the answers consistent, correct and specific? (c) sameAs coherence audit: an automated crawl across every declared third-party profile, comparing attributes against the master declaration — deviations are reported as breaks. (d) Citation rate in the own field — 200–500 field-specific prompts across a multi-model tracker, measuring author citation over time.

These four metrics together form the author-entity strength score, which we track as a sub-metric for subject-matter authors as standard inside LLM Citation Monitoring.

Conclusion: authors are assets, not attributes

The structural shift of the last three years is not that E-E-A-T became more important — it always was. The shift is that E-E-A-T evidence has become machine-readably testable. Google can read author information from schema, LLMs read them from structured data and entity graphs, AIO draws from both. Brands that treat authors as attributes of articles build content infrastructure on a fragile foundation. Brands that understand authors as standalone, long-term entity assets accumulate E-E-A-T signals that radiate across the entire content output and compound across classical search and LLM citations alike.

The operational consequence: author-entity build-out belongs as a dedicated workstream in every serious SEO program. Not as an SEO detail, but as a strategic investment with a three-year return. The effort per author is clearly quantifiable — 120 days of structured work plus quarterly maintenance. The return is in higher YMYL rankings, higher AIO citation rate, consistent LLM citation, and a brand whose subject-matter expertise is machine-readably verifiable.

Sources

• Google Search Central: Creating helpful, reliable, people-first content (last updated December 10, 2025)
• Google: Search Quality Rater Guidelines (PDF, version of September 11, 2025)
• Google Search Central Blog: Our latest update to the quality rater guidelines: E-A-T gets an extra E for Experience (December 15, 2022)
• Google Search Central: Article structured data, author markup best practices (accessed September 23, 2026)
• Google (Danny Sullivan): Google’s Knowledge Graph and Knowledge Panels (May 20, 2020)
• Wikidata: Wikidata:Notability (accessed September 23, 2026)
• Wikidata: Wikidata:Licensing (accessed September 23, 2026)