E-E-A-T

What the four signals are

E-E-A-T is shorthand for four qualities a source can show: demonstrated experience, subject-matter expertise, authoritativeness and trustworthiness. The AI Overviews article describes them as the signals search engines have described for years when judging whether a page deserves trust. The citation article uses the same shorthand for credible sourcing.

The four are not one score. The AI Overviews article ties them to things a page can show: clear authorship, evidence of direct experience of the subject, and a track record a model can find corroborated elsewhere. A page can show some of these and not others, which is why the term works as a checklist and not only as a label.

Why AI Answers lean on it

A search engine ranking page and a generative model choosing what to cite are answering related but different questions. Ranking has long weighed relevance and authority together. A model assembling a generated answer has an extra, narrower filter on top: it chooses which handful of sources to attribute a claim to, in an answer it presents as reliable. The AI Overviews article calls that closer to an editorial decision than a ranking decision.

That is why the same article says authority signals matter more under this model, not less. Being selected as a trustworthy source is a precondition for being cited at all. A page can be technically well optimised and still not read, to a model deciding what to cite, as a source worth attributing a claim to. Being well written is necessary but no longer sufficient. Being credibly the kind of source an answer should be attributed to is the harder bar underneath it.

There is a second reason it matters. A model that cannot find a trustworthy source still produces an answer, and the answer can read as confident while stating something wrong. That is a hallucination as it looks from outside. An answer that is properly grounded points back to a real, checkable source for each claim, and a source with weak signals is a poor thing to point back to.

What the signals look like on a page

The AI Overviews article is specific about this. Clear authorship is the first signal: a named person responsible for the page. Evidence of direct experience is the second: the page shows the author has done the work, with detail a summary of other pages could not supply. A track record a model can find corroborated elsewhere is the third.

The brand article makes the same point from the other side. Experience, expertise, authority and trust are judged from visible evidence: named people, real credentials, clear provenance. It also asks that the plain facts about a business appear on its own site in ordinary sentences, covering what it does, who for, where it operates and who is behind it. An assistant, like a new customer, needs a sentence it can lift and repeat accurately.

None of that is hidden in code. It is written on the page for a person to read, and a model reads the same words.

The Evidence outside your own site

Authority is not only what a site says about itself. The brand article notes that independent reviews, forums and articles tend to be given more weight than a company’s own pages, because a company website is read as promotional. A strong set of facts on your own site, important as it is, does not settle the matter. What independent sources say about you is part of the evidence a model weighs.

The citation article gives a figure that points the same way. Brand mentions correlate roughly three times more strongly with AI visibility than backlinks do, measured at 0.664 for mentions against 0.218 for backlinks. Earning a mention across other sites and publications, even without a link attached, appears to carry weight that traditional link-building metrics do not capture.

Consistency is the other half. The brand article explains that when a site says one thing, a directory says another and an article says a third, the assistant has to choose or blend, and a blend produces a plausible but wrong answer. Its advice is to make every listing say the same thing and to earn accurate mentions on independent sites. It adds that this cannot be shortcut, and that attempts to fake it create exactly the conflicting signals you are trying to remove.

Techniques that make a Source easier to Trust

The study in the citation article tested nine content techniques, and the three with the largest gains all add verifiable, attributable information. Adding statistics improved citation visibility by 41%. Adding direct quotations improved it by 28%. Citing external sources improved it by 115% for lower-ranked content. The article reads these as evidence that AI systems prefer content offering something specific and checkable over general assertion.

That sits comfortably beside the four signals. A named statistic with a source shows trustworthiness. A quotation from a named expert shows that expertise exists somewhere in the chain. A cited source lets a reader and a model check the claim. The same study found that keyword stuffing produced no meaningful benefit and that adding more words did nothing on its own. The signal is data density and source credibility, not length.

Authority is half the job

The AI Overviews article is clear that clearing the authority bar is only half the work. A separate, more mechanical question decides whether a source gets used: can the model extract a clean, self-contained answer from it without much interpretation. A direct answer near the top of a section, a clearly labelled list and specific facts in sentences that stand alone are easier to lift than the same information buried in a long paragraph.

The article adds that content freshness and factual specificity sit alongside structure. A model choosing between two structurally similar sources leans towards the one with a checkable date and a specific figure over the one making the same claim in vaguer terms. So authority gets a page into the running, and structure and specificity decide whether it is used.

What Google says to focus on

The SEO against GEO review reads Google’s guide to its generative AI features, last updated in July 2026. The guide says optimising for generative AI search is still SEO. It names tactics that can be skipped: new machine-readable files, AI text files, special markup, splitting pages into small chunks, and manufactured brand mentions. What it recommends instead is unusual, non-commodity content with first-hand expertise, a sound technical foundation and accurate business data.

Read against the four signals, the recommendation is close to a restatement of them. First-hand expertise is experience and expertise. Accurate business data is trustworthiness. The review adds that the guide covers Google only, and that other assistants retrieve and cite differently. A 15,000-query study it cites found that on average only 12% of AI citations came from pages in Google’s top 10. The signals are the safer thing to build on because they do not depend on any one system’s quirks.

Mix-ups worth avoiding

Four confusions come up often enough to name. The first is treating E-E-A-T as markup. The review reports Google saying there is no special schema.org markup to add for its AI features, so there is nothing to switch on. The signals are visible content: people, evidence and sourcing.

The second is treating authority as a synonym for size. The citation article says the opposite for citations: lower-ranked and newer sites stand to gain the most from the techniques it measured, because the overlap between rankings and citations has fallen sharply.

The third is assuming the signals only matter to Google. They matter wherever a model has to choose which few sources to attribute an answer to, which includes ChatGPT, Gemini, Copilot and Perplexity as the AI citation page describes them. The fourth is assuming strong signals guarantee a citation. They do not. They put a page in the running.

Who has most to gain

The review notes that deliberate GEO work fits businesses whose expertise is the product, such as consultancies, specialist services and technical suppliers, because original knowledge is exactly what assistants quote. That is where the four signals are easiest to show truthfully. A firm that has done the work can say who did it, what was learned and where the evidence is.

The effort the review describes moves from phrasing to substance: original facts, named expertise and clear sourcing. That is good news for a small team, because those are things it can produce without writing more. The citation article makes the same point about length. The techniques that move the needle are additions to existing content, not a requirement to produce a great deal more of it.

How to check where you stand

The brand article gives a method that needs no special software. Write a fixed set of ten to twenty-five questions in the words a buyer would use, run them in clean sessions across the assistants you care about, and record whether you were named, what was said about you, whether each claim was accurate, outdated or wrong, and which sources were cited. Repeat the identical set monthly, because one run proves nothing and the pattern across runs is the real signal.

Read the results with the four signals in mind. If you are absent, the assistant may not have enough clear, consistent material to recognise you. If a claim is wrong, trace it to an old or conflicting source and correct that source. If you are framed badly, look at the comparison pieces that group you with others. The ranking article adds that citation should be tracked as its own measure, apart from rank position.

Where it fits in an Audit

Being in the running to be cited is a separate question from what gets content cited once it is, and a business that gets the first wrong never reaches the second. That is what an AI audit checks when it looks at a site’s exposure to generative AI search. It starts with a discovery call and follows the five steps in how we audit. Most of the work is not a rewrite. It is a review of what a site already says about who is behind it, what they have done and what can be checked. To have a site checked directly, see get an audit.

Related reading

The reasoning is set out in how AI Overviews are changing what counts as SEO. The techniques that move citations are in what gets content cited in AI answers. Checking what assistants say about a business is covered in does your brand show up when people ask AI about you, and the wider comparison is in the SEO against GEO review. The practice that puts the signals to work is generative engine optimisation, and the discipline the term comes from is search engine optimisation.