Findings · 2026-09-01 · By VSNARY | Emmanuel Orta · 0 views

E-E-A-T is not on the wire. Here is what a crawler can measure instead.

Experience, expertise, authoritativeness and trust are judgements an evaluator makes. A fetch cannot make them. It can check whether the evidence for them exists in a form a machine can find.

Every third SEO tool now has an E-E-A-T score. It is worth being precise about what such a score can be. Experience, expertise, authoritativeness and trust are qualities a human evaluator judges after reading, comparing, and knowing things about the world. None of that is in an HTTP response. A tool that claims to score E-E-A-T from a fetch is scoring something else and naming it after the thing you wanted.

What is in the response is a set of statements the site makes about itself in a form a reader could use as evidence. Whether a named person is declared. Whether that person carries a job title, a credential, a membership, a field of knowledge. Whether any profile elsewhere corroborates them. Whether articles name a typed author and carry both dates. Whether the organisation says who runs it and when it was founded. Whether the pages an evaluator opens first, about, contact, policy, are linked from the homepage at all. Each is measurable. None of them is expertise.

Fourteen rows, and what they are careful not to say #

The section that shipped today scores those statements from the delivered homepage and its JSON-LD. Each row states what it read, what the target is, and why it matters; the block underneath says plainly that a full row of passes does not create expertise and a row of fails does not disprove it. What the rows decide is narrower and more useful: whether a machine reading this site can find the evidence at all. A person who exists only as a name in a script tag is anonymous to a resolver, however qualified they are in life.

The rows are written so they cannot accuse. A site that declares no person reads unmeasured on the expertise and corroboration rows, not failed. The check does not read expertise from microformats, so a person declared only as an h-card counts as named and the rest stays unmeasured. And a FAQ page is not an article: on the first run, two sites were told their FAQ had no author and no dates. A FAQ has neither by convention, and that is corrected.

Measured on ourselves, and what it found #

This site failed five of the rows on the first pass. The Organization node named nobody and carried no founding date. The founder's Person node had no job title and no field of knowledge. The homepage linked no about page and no contact route. All five were true, all five were fixed the same afternoon, and one of the fixes exposed a new row's blind spot: the about page here is /author, which the check did not recognise. An author page answers the same question an about page does, and it now counts.

The founder reference was the interesting one. Adding founder: {"@id": "/#author"} to the Organization node only helps if that @id is declared on the same page; otherwise it is a dangling reference, the exact defect the new RDF section counts against everyone else. So the Person node is now emitted on every page, not just the author page. Declaring a relation and declaring its target are two statements, and a graph needs both.

The judgement, and what a response can actually carry #

Each is measurable. None of them is expertise. A site with no Person node reads unmeasured, not failed.

What to publish, in order #

A named Person with a jobTitle and a url. A sameAs to one profile that links back. On the organisation, founder pointing at that person and a foundingDate you can defend. On every article, author as an object, not a string, and both datePublished and dateModified. About, contact and policy linked from the homepage. That is the whole list, it is a morning's work on most sites, and after it a machine can at least find the evidence. Judging it is still someone else's job. The entity SEO guide covers where those statements sit in the wider graph; the glossary defines E-E-A-T proxies as distinct from E-E-A-T.

Questions this post answers

Can a tool score E-E-A-T from a page fetch?

Not honestly. Experience, expertise, authoritativeness and trust are judgements an evaluator makes about the world. A fetch can measure whether the machine-readable evidence for them exists: a named person, expertise properties, corroborating profiles, typed authors and dates, an organisation that names its founder, and linked about, contact and policy pages.

What should I publish first?

A Person node with a name, jobTitle and url; a sameAs to one profile that links back; founder and foundingDate on the organisation; author as an object plus datePublished and dateModified on every article; about, contact and policy linked from the homepage.

Related findings

Comments

Comments are read before they appear. Nothing is published automatically, and no account is needed.

Writing about this? Facts, live figures and marks — every number on that page is dated and traceable to a scan.

All findings · The dataset · How the dataset works