E-E-A-T Is a Description, Not a Strategy: Credibility Needs a Diagnostic
The SEO industry has spent the better part of a decade optimising for E-E-A-T, auditing for E-E-A-T, selling E-E-A-T packages, running E-E-A-T webinars, and writing E-E-A-T checklists, and almost nobody selling any of it has stopped to ask the awkward question underneath: optimising it for what, exactly? E-E-A-T has no score, no threshold, no lever, and no defined intervention, and the document it comes from was never written for you in the first place.
E-E-A-T was written for Google’s raters, not for you
Read the provenance before you read the acronym. E-E-A-T comes from the Search Quality Rater Guidelines, a document Google wrote to give its human quality raters a shared vocabulary for evaluating content, and Google has said plainly, repeatedly, that it isn’t a direct ranking factor and there’s no E-E-A-T score inside the algorithm. That’s an evaluator’s checklist, written for evaluators, doing evaluator work. The industry picked it up, read it as an operating manual for entities, and built a service economy on the misreading. None of this is Google’s doing: Google wrote a rater document, said what it was for, and kept saying it. The misreading is ours, and the industry had every incentive to keep misreading, because you’ve been handed the exam marker’s rubric and told it’s the syllabus by the people selling syllabus tuition.
A description tells you what good looks like, a diagnostic tells you what to do next
Here’s the structural problem, and it’s the reason E-E-A-T keeps you busy without getting you anywhere: E-E-A-T describes, it doesn’t diagnose. Experience, Expertise, Authoritativeness, Trustworthiness: four properties that matter, listed flat, with no indication of whose judgement makes each one real or which behaviour moves it. A description tells you what good looks like, a diagnostic tells you whose behaviour must change and through which lever, and effort spent optimising against a description with no diagnostic structure underneath is effort without a target. That’s how a useful vocabulary becomes a shiny object: nobody designed the distraction, the industry built it itself, because a named construct is easier to sell than a diagnosis, and it cannot tell a single brand what to do on Monday morning.
I added Notability because E-E-A-T had no cell for famousness
For me, the tell was always what the list left out. Years ago I extended E-E-A-T into NEEATT, adding Notability, with Transparency contributed by Jarno van Driel, because the four properties had no cell for recognition: no way to represent the difference between being known and being deferred to. A framework that can’t represent famousness can’t diagnose the most expensive confusion in the game, and the confusion has a pedigree: popularity read as authority is the PageRank pitfall, the original sin of link-based ranking, where being pointed at and being trusted got measured as the same signal. Twenty-five years later, a flat list of trust adjectives repeats the conflation at the framework level, and the brands paying for E-E-A-T audits inherit it.
Nine properties break Credibility into implementable chunks
So I finished the job. The model is now formalised, deposited, and open access: Trust as Triangulation, a three-evaluator, nine-property credibility model. Three evaluator positions, Self, Peers, and Audience, each constituting three properties, and every one of the nine cells carries a definition, a behaviour it measures, an evidence type, and an improvement lever. That last part is the whole point: implementable chunks. The peer axis alone splits the old “Authoritativeness” blob into three behaviours that move independently, deference (Authority), reference (Notability), and agreement (Corroboration), and the moment you can see them separately, you can see which one is your bottleneck and whose behaviour has to change. E-E-A-T maps to four of the nine cells, NEEATT to six. The grid does what the checklist never could: it tells you what to do next.
Whatever framework you prefer, it’s all Credibility in UCD
And here’s the part where I genuinely don’t mind if you ignore my grid: build your own. Eleven cells, seven cells, your agency’s house model, knock yourself out, because every one of them is a decomposition of the same layer. The layer is Credibility, the middle layer of UCD (Understandability, Credibility, Deliverability): does the machine believe you’re good at what you do. E-E-A-T is a credibility instrument, NEEATT is a credibility instrument, my nine-cell grid is a credibility instrument, and so is whatever gets invented next, which means the durable question was never “how’s your E-E-A-T”, it was always “how’s your Credibility, and have you built the Understandability underneath it for the Credibility to attach to”. Stop polishing the rubric, start building the layer: the machine was never grading your acronym.
Publication note: the argument in this article is formalised in the deposited academic paper Trust as Triangulation: From E-E-A-T to a Three-Evaluator, Nine-Property Credibility Model for Human and AI-Mediated Evaluation, which carries the evaluator-specificity claim, the nine-cell grid, the E-E-A-T and NEEATT mappings, and the credibility category claim with full references. NEEATT: Notability added by Jason Barnard, Transparency contributed by Jarno van Driel. The UCD build-order argument is developed in the companion Sandbox piece, Marketing Builds Backwards. Jason Barnard, 13 June 2026.