E-E-A-T in SEO: Trust Signals for Google and AI Search in 2026

What Is E-E-A-T? Experience, Expertise, Authoritativeness and Trust

E-E-A-T (also written EEAT) stands for Experience, Expertise, Authoritativeness and Trust: the qualities Google's search quality raters look for when they judge whether a page, and the people behind it, can be trusted on a topic.

Search quality raters are people Google pays to rate search results, so it can check how well its search systems work. Their ratings don't set rankings.

E-E-A-T is Google's name for what makes a page worth trusting: first-hand experience, real expertise, a reputation as a go-to source, and above all Trust itself. It describes evidence that Google's human raters look for, not a score a page earns or a setting inside the ranking algorithm.

Google publishes the raters' handbook, the Search Quality Rater Guidelines, as a public PDF. The version current in September 2026 is dated September 11, 2025, and its section 3.4 defines the four members:

Member	What the rater guidelines mean	Example
Experience	"the necessary first-hand or life experience for the topic"	Reviews of tax software from people who have used it (Google's 2022 example)
Expertise	"the necessary knowledge or skill for the topic"	Guidance on filing a tax return from someone with tax expertise (Google's 2022 example)
Authoritativeness	"known as a go-to source for the topic"	A site others in the field point to on that subject (our example)
Trust	"the extent to which the page is accurate, honest, safe, and reliable"	An online store with secure payment and reliable customer service (the guidelines' example)

The guidelines put one member above the others: "Trust is the most important member of the E-E-A-T family because untrustworthy pages have low E-E-A-T no matter how Experienced, Expert, or Authoritative they may seem."

Google added Experience on December 15, 2022, turning E-A-T (Expertise, Authoritativeness, Trustworthiness) into E-E-A-T. Older articles still use the E-A-T name for the same idea, minus Experience.

Trust signals, in SEO, are the checkable evidence behind E-E-A-T: who wrote a page, what independent sources say about the site, whether its claims are sourced.

The same phrase also means the badges and security seals used to raise conversions on a checkout page. Where the two meanings overlap, as with secure payment, customer service and honest reviews, the rater guidelines list them too.

Is E-E-A-T a Ranking Factor? What Google Has Said, Side by Side

E-E-A-T isn't a ranking factor or a score: it's a standard Google's human raters apply, and Google says its ranking systems use other signals that tend to identify content with good E-E-A-T. Google's own statements, from different sources and years:

Source	Type and date	What it says
Creating helpful, reliable, people-first content	Google documentation, updated December 2025	"While E-E-A-T itself isn't a specific ranking factor, using a mix of factors that can identify content with good E-E-A-T is useful."
Same page	Google documentation, updated December 2025	"Search raters have no control over how pages rank. Rater data is not used directly in our ranking algorithms."
Search Quality Rater Guidelines, section 0.1	Guidance for raters, September 2025	"No single rating can directly impact how a particular webpage, website, or result appears in Google Search"
Danny Sullivan, then Google's Search Liaison, as reported by SE Roundtable	A post on X, not documentation, February 7, 2024	"It's not a ranking factor. It's not a thing that's going to factor into other factors."
"How Google Fights Disinformation"	Google white paper, February 2019, before Experience was added	Algorithms "identify signals about pages that correlate with trustworthiness and authoritativeness"; PageRank, built on links, is the best known
A guide to Google Search ranking systems	Google documentation, updated December 2025	Lists reliable information and reviews systems, among others; helpful content is listed as retired, part of core ranking since March 2024. No "E-E-A-T system"

The statements don't line up neatly. The help page says a mix of factors can identify content with good E-E-A-T; Sullivan's post says E-E-A-T isn't a thing that factors into other factors.

Read together, and this is our interpretation, neither describes a system that measures E-E-A-T as such. What Google's systems measure are other signals, such as links, reviews and reliability signals, that correlate with what raters would call high E-E-A-T.

Correlate means they tend to rise and fall together, which doesn't show one causes the other.

The practical consequence is that there is no E-E-A-T number to raise. A tool that reports an "E-E-A-T score" is producing its own estimate, not Google's.

E-E-A-T is evidence, not a score. The question is whether real, checkable evidence shows a page deserves trust: Google's human raters judge it directly, and its ranking systems only through signals that tend to go with it. The work is building that evidence, not chasing a number or a checklist item.

The topic changes how much that evidence counts. The same help page says Google's systems "give even more weight to content that aligns with strong E-E-A-T" on topics that could significantly affect people's health, finances or safety, or society's welfare. Google's raters call those YMYL topics.

What Is YMYL? Why Health, Money and Safety Topics Need Stronger Trust Signals

YMYL (Your Money or Your Life) topics are those that "could significantly impact the health, financial stability, or safety of people, or the welfare or well-being of society", in the rater guidelines' definition (section 2.3). Pages on them need stronger trust signals than other pages.

The guidelines' YMYL types, with example topics from the guidelines' own list:

• YMYL Health or Safety: "symptoms of a heart attack", "what to do if there is an earthquake".
• YMYL Financial Security: "how to invest money".
• YMYL Government, Civics & Society: "who can vote".
• YMYL Other: topics that could hurt people or society in other ways.

Beyond the extra weight in Google's help page, two more Google documents raise the bar. Google's guidance on AI-generated content says that for YMYL topics, "our systems place an even greater emphasis on signals of reliability".

The rater guidelines add that "the reputation of a website should be judged by what experts in the field have to say" (3.3.1).

Experience can be enough on a YMYL topic, but some advice needs expertise. Section 3.4.1 says first-hand life experience on a YMYL topic "may be considered to have high E-E-A-T as long as the content is trustworthy, safe, and consistent with well-established expert consensus".

The same section adds that "some types of YMYL information and advice must come from experts". As an illustration of our own: someone describing life with a chronic condition can show real Experience, while advice on medication doses has to come from a medical expert.

The bar also drops where the stakes are low. Google's help page says of the four members: "The others contribute to trust, but content doesn't necessarily have to demonstrate all of them."

The higher the stakes of a topic, the more trust evidence a page on it needs. On health, money, safety and civic topics, lived experience can still count, but some advice has to come from experts. A recipe page and a medication page don't need the same evidence.

Where Trust Signals Come From: What a Site Says, What Others Say, What's on the Page

Google's rater guidelines look for trust evidence in three places: what a site says about itself, what others say about it, and what's visible on the page.

Section 3.4 names the three as "What the website or content creators say about themselves", "What others say about the website or content creators", and "What is visible on the page, including the Main Content and sections such as reviews and comments".

Reviews show how to tell the three apart. Reviews displayed on a site's own page are what's on the page; reviews collected by independent sites are what others say.

Who produced a piece of trust evidence decides what it's worth: the site itself, an independent third party, or the page's own checkable content. Claims a site makes about itself are the easiest to produce, so they count for little until independent sources or the page itself back them up.

What a Site Says About Itself: Bylines, Author Pages and Who Is Responsible

What a site says about itself is who wrote each page, who stands behind the site and how to reach them. It's the easiest trust evidence to add, and the easiest to fake.

Google's help page frames it as three questions:

• Who created the content: "Do pages carry a byline, where one might be expected?" Google says it "strongly encourage[s] adding accurate authorship information".
• How it was created: how the content was produced and, for AI-assisted content, a disclosure, covered below.
• Why it was created: "'Why' is perhaps the most important question", and the right answer is to help people, not to attract search visits.

The signals themselves:

• A byline, the author's name shown on the page, wherever readers expect one.
• An author page, a page about the writer, linked from the byline, with real credentials and experience. The help page asks for "background about the author or the site that publishes it, such as through links to an author page or a site's About page".
• An About page that says who owns and runs the site.
• Contact and customer service information. The guidelines say it "should be clear" who is responsible for the site and who created the content (2.5.2), while accepting aliases on forums and social platforms.

For pages that need a high level of trust, the guidelines call information about who created the content and who is responsible for it "critical" (4.5.1).

A bio doesn't add trust on its own. Danny Sullivan's February 2024 posts on X, as reported, put it bluntly: "Having an expert write things doesn't magically make you rank better", since anyone can declare themselves an expert.

The rater guidelines say the same from the rater's side: "Be skeptical of claims that websites make about themselves, particularly when there is a clear conflict of interest." It follows, in our reading, that a credential counts when others can confirm it.

Inventing credentials is worse than having none. The guidelines rate false credential claims and made-up creator profiles Lowest, their bottom rating (4.5.3).

Marking up the authorship a page already shows, so machines can read it, is the subject of the Structured Data & AI Entity Optimization guide.

What Others Say: Reputation, Reviews and Independent Mentions

What others say is the reputation evidence a site can't write for itself: independent reviews, news coverage, and references from experts and encyclopedias.

Google's help page turns it into a test: "If someone researched the site producing the content, would they come away with an impression that it is well-trusted or widely-recognized as an authority on its topic?"

The signals:

• Independent reviews of the business.
• News articles about it.
• References from recognized experts or institutions.
• An encyclopedia entry, such as Wikipedia, which the guidelines tell raters to search (3.3.3).
• For academic authors, peer-reviewed publications (3.4).

Customer reviews count, read with care. The guidelines tell raters to "Be skeptical of both positive and negative reviews", and say "the content of the reviews matter, not just the number or star rating" (3.3.2).

Independence is the test for all of it. A conflict of interest is when a source gains from what it says. The guidelines say manufacturer "reviews" and paid influencer "reviews" "are not as trustworthy due to the conflict of interest" (3.4).

Vendor link scores aren't Authoritativeness. Moz's Domain Authority and Ahrefs' Domain Rating are scores SEO tools calculate from links. The guidelines define Authoritativeness as being "known as a go-to source for the topic".

That definition makes authority topic by topic: a site can be the go-to source on one subject and unknown on the next. Building depth on a topic through site structure is the subject of Site Architecture & Topical Authority for AI.

The guidelines also say traffic data such as Similarweb's "do not provide evidence of positive or negative reputation" (3.3.1). A third-party metric describes a site; it isn't what people say about it.

Faking this evidence doesn't work either. Google's guide to generative AI features says "seeking inauthentic 'mentions' across the web isn't as helpful as it might seem."

We'd argue that's what makes this source the most telling of the three: it's the one a site controls least. Building it is offsite work, with guides of its own:

• How links, PR and third-party presence fit together is the subject of Offsite SEO Strategy.
• Earning links and press coverage goes deeper in Backlinks & Digital PR.
• Being present on the review sites, forums and communities that AI answers pull from through retrieval-augmented generation (RAG) is the subject of Third-Party Presence & RAG Visibility.
• Reviews on a Google Business Profile belong to Local SEO for AI Search.

What's on the Page: Sources, First-Hand Evidence and Accuracy

What's on the page is the evidence a reader can check without leaving it: sources for its claims, proof of first-hand experience, and facts that hold up.

• Claims linked to their sources. The help page asks for "clear sourcing, evidence of the expertise involved".
• First-hand evidence wherever experience is claimed: photos, test details, measurements.
• No "easily-verified factual errors", in the help page's words.
• Conflicts of interest disclosed, such as affiliate links or reviews of the site's own products. That's our extension of the guidelines' conflict-of-interest point, not a line from Google.
• Reviews and comments shown on the page, which the guidelines count as part of what's visible.

The guidelines don't rank this source against the other two, but they note that "for some types of pages, the level of experience and expertise may be clear from the MC itself", MC being the page's main content. Their example is a video of someone styling hair.

Review content has its own standard. Google's advice to write high quality reviews says: "Provide evidence such as visuals, audio, or other links of your own experience with what you are reviewing, to support your expertise and reinforce the authenticity of your review."

The help page's "How" question asks the same of product reviews: how many products were tested, and "evidence of the work involved, such as photographs". Google's reviews system "aims to better reward high quality reviews", including those "written by experts or enthusiasts who know the topic well".

For online stores, the guidelines' own Trust example, "secure online payment systems and reliable customer service", is on-page evidence too. Whether the connection is secure (HTTPS) is a technical check, part of the floor the Technical SEO guide sets out.

AI-Written Content and E-E-A-T: Disclose How It Was Made, Never Invent an Author

Google doesn't penalize content for being written with AI. It rates low-effort content and deceptive authorship low, and it asks sites to say how content was made rather than give the AI a byline.

Google's February 2023 guidance on AI-generated content says "Appropriate use of AI or automation is not against our guidelines." It also says: "Using AI doesn't give content any special gains. It's just content."

Effort is what the rater guidelines judge. AI-generated main content made "with little to no effort, little to no originality, and little to no added value" is rated Lowest, and "the use of Generative AI tools alone does not determine the level of effort or Page Quality rating" (4.6.6), the raters' overall score for a page.

At volume, the same pattern is scaled content abuse in Google's spam policies: when many pages are generated "for the primary purpose of manipulating search rankings and not helping users." Producing and reviewing AI-assisted content at scale is the subject of AI-Assisted SEO Content Production.

What's left is authorship. Google's documents describe three patterns, and a fourth follows from them:

Pattern	What Google says	Source
A named person accountable for the page, with AI use disclosed where readers would ask	Accurate bylines where someone might think "Who wrote this?"; AI disclosures where someone might think "How was this created?"	AI-content guidance; help page
The AI tool credited as the author	"Giving AI an author byline is probably not the best way to follow our recommendation to make clear to readers when AI is part of the content creation process."	AI-content guidance
A made-up human author for AI content	"'fake' owner or content creator profiles", such as "AI generated content with made up 'author' profiles ...": deceptive, rated Lowest	Rater guidelines, 4.5.3
No disclosure where readers would reasonably ask how the page was made	Falls short of the "How" recommendation (our reading, not a Google statement)	Help page

Drafting with AI tools or an agent is a question of who does the work, which the Agentic SEO guide covers. For trust, what matters is that a named person stands behind the page and the way it was made is disclosed.

Does E-E-A-T Carry into AI Answers? What Google, Bing and OpenAI Document

For Google's AI features, yes in effect: Google says AI Overviews and AI Mode run on the same core ranking and quality systems as Search. Bing states the same foundation in general terms, and OpenAI documents only how ChatGPT's crawler gets access, not how it judges trust.

AI answers that cite pages rely on grounding: the AI looks up pages at the moment of the question and builds its answer from them, as the AI SEO guide explains. The open question is what each engine documents about trust when it picks those pages.

The evidence differs in strength, so each row carries a label: documented by the engine, described in general terms, access rules only, research on a benchmark, vendor correlation, or no source. Checked September 2026:

Engine or source	What it says about trust	Evidence tier
Google AI Overviews and AI Mode, guide to generative AI features	Rooted "in our core Search ranking and quality systems"	Documented by the engine
Google, AI features and your website	"There are no additional requirements to appear in AI Overviews or AI Mode, nor other special optimizations necessary"	Documented by the engine
Google AI Overviews, 2024	"AI Overviews are built to only show information that is backed up by top web results" (Elizabeth Reid, May 2024)	Documented by the engine
Microsoft Bing and Copilot, Bing Webmaster Guidelines section 1	"Bing and Copilot search experiences rely on the same core crawling, indexing, and ranking foundation as traditional search"	Described in general terms
Bing Webmaster Guidelines, sections 1, 11, 15	"Authority and trust signals" (1); "Be original and authoritative" (11); "Ensure Content can be Verified Independently" (15). No author or reputation criteria	Described in general terms
ChatGPT search	"Any public website can appear in ChatGPT search", provided OpenAI's search crawler, OAI-SearchBot, isn't blocked (OpenAI's publisher FAQ)	Access rules only
The generative engine optimization (GEO) paper, Aggarwal et al., KDD 2024	Citing sources, quotations and statistics: "a relative improvement of 30-40%". An authoritative tone: "no significant improvement"	Research on a benchmark
Ahrefs, 75,000 brands, December 2025	Branded web mentions correlate with AI visibility (about 0.66 to 0.71) more than Domain Rating does (about 0.27 to 0.33); "correlation isn't causation"	Vendor correlation
Circulating E-E-A-T and AI-citation percentages; a claimed "June 2026" rater guidelines update on AI authorship	No published method; the official PDF is still dated September 11, 2025, and the cited sections carry other titles	No source

It follows, though Google doesn't put it this way, that the trust evidence helping a page in classic search is, for Google, the evidence that gets it used in AI answers. Google hasn't documented E-E-A-T as deciding AI citations.

Beyond Google, only the benchmark and the vendor data speak to this, and they point the same way: toward sources and independent mentions over sounding authoritative. The GEO result comes from a simulated engine and the Ahrefs figures are correlations, so neither proves it.

No engine documents E-E-A-T as a "filter" or "gate" for AI citations, although the claim circulates. Google documents the same core systems and no additional requirements.

Google has also shown where experience and reliability pull apart. After AI Overview errors in May 2024, it "updated our systems to limit the use of user-generated content in responses that could offer misleading advice", while calling forums "often a great source of authentic, first-hand information".

Vendor numbers on AI citations age fast. Ahrefs reported in July 2025 that 76% of AI Overview citations came from the top 10 results, and in March 2026 put it at 38%. Ahrefs attributes part of the change to Google moving AI Overviews to Gemini 3 in January 2026, and part to its own improved parsing.

The table's last row carries the simplest lesson: check the primary document before repeating a claim.

Google says its AI answers run on the same core systems as Search, so in practice the same trust evidence counts for both. Other AI search engines don't publish their trust criteria. The best outside evidence favors citing sources and being mentioned by independent sites over sounding authoritative.

How each engine picks its sources goes deeper in How AI Search Engines Work, and tracking whether a site gets cited is the subject of AI Visibility, Mentions & Prompt Monitoring.

An E-E-A-T Audit: Checking the Trust Evidence a Site Already Has

An E-E-A-T audit checks the trust evidence in the three places raters look, page by page and for the site and its authors, starting with the pages where trust matters most. It ends with a list of gaps and who fixes each.

Some steps are data work, with documented, repeatable inputs and outputs: running set searches, listing pages, checking whether a byline or source link is present. Others are judgment calls: whether a source is independent, whether a credential is real and relevant, whether a page is YMYL.

The split and the step order are our proposed method, not a published one. The order runs from how much trust each page needs, through the three evidence sources, to fixes.

1. Sort Pages by How Much Trust They Need

• In: the site's page list.
• How: mark each page YMYL or not, using the guidelines' definition, and mark the pages that carry the business.
• Out and done: every page has a trust bar: YMYL, business-critical or standard.
• Type: both. Listing pages is data work; deciding what's YMYL is a judgment call.

2. Search What Others Say About the Site and Its Authors

• In: the brand name, the domain and each author's name.
• How: run the reputation searches in section 3.3.3 of the rater guidelines, which are Google's own instructions to its raters. Google's example uses IBM; the searches below swap in a placeholder brand.
• Searches: [brand -site:brand.com], ["brand.com" -site:brand.com], [brand reviews -site:brand.com], ["brand.com" reviews -site:brand.com] and [brand site:en.wikipedia.org].
• Authors: repeat with each author's name plus their topic. The guidelines apply reputation research to content creators too; adding the topic is our suggestion.
• Log: each independent source, what it says, whether it's positive or negative, and whether it's truly independent. For YMYL pages, note what experts in the field say.
• Out and done: a reputation log for the brand and for each author.
• Type: both. Running the searches and logging results is data work; judging independence and what a source means for reputation is a judgment call.

3. Check What the Site Says About Itself

• In: the pages from step 1, plus the About, contact and author pages.
• How: check for a byline where readers expect one, a byline that links to an author page, credentials and experience the step 2 log confirms, an About page that says who is responsible, reachable contact and customer service, and secure payment on stores.
• Out and done: per page, each item marked present or missing, plus any claim no outside source confirms.
• Type: both. Presence is data work; whether a credential is real and relevant is a judgment call.

4. Check What's on the Page

• In: the same pages.
• How: check that claims link to sources, primary where possible; that claimed experience shows evidence such as photos or test details; that there are no easily checked factual errors; and that reviews and comments shown are real.
• Out and done: per page, a list of unsourced claims, missing evidence and errors.
• Type: both. Finding claims without links is data work; judging accuracy and evidence is a judgment call.

5. Flag Trust Breakers

• In: the results of steps 2 to 4.
• How: look for undisclosed conflicts of interest, credentials or claims with no outside confirmation, invented or inflated author profiles, and AI-assisted content with no disclosure where readers would ask how it was made.
• Out and done: a list of trust breakers, each with its page and its fix.
• Type: judgment call.

6. Turn Gaps into a List with Owners

• In: the results of steps 1 to 5.
• How: give each gap exactly one owner from the list below.
• Owners: an on-page or on-site fix; markup for authorship the page already shows (Structured Data & AI Entity Optimization); reputation, links and mentions (Offsite SEO Strategy, Backlinks & Digital PR, Third-Party Presence & RAG Visibility).
• Or research: a topic the site can't credibly cover yet is a research decision, which SEO Content Fundamentals covers.
• Priority: trust breakers first, then YMYL and business-critical pages, then the rest. That order is our judgment, not a documented rule.
• Out and done: every gap has an owner and a priority.
• Type: judgment call.

Every value in the row below is illustrative:

Page	Evidence source	Gap	Owner	Priority
A medication dosage page	What the site says about itself	No byline or author page	On-site fix	High: YMYL

The rule under every step is that the audit looks for evidence others can check, not boxes ticked. A byline whose credential nobody outside confirms, or a review with no sign of testing, is still a gap.

Running this audit on a real site, with its own pages, authors and search results, is separate work.

Sources

Sourced claims:

• Search Quality Rater Guidelines, Google, September 11, 2025. Guidance for human raters, not the ranking algorithm. Cited for the four members, Trust as most important, YMYL and its types, the three evidence sources and conflict of interest.
• The same guidelines, also cited for: experience vs experts on YMYL (3.4.1), responsibility (2.5.2, 4.5.1), skepticism and expert reputation (3.3.1), customer reviews (3.3.2), the reputation searches (3.3.3), fake profiles (4.5.3), low-effort AI content (4.6.6), the on-page example (3.4) and "No single rating" (0.1).
• Creating helpful, reliable, people-first content, Google Search Central (updated December 10, 2025). Cited for "isn't a specific ranking factor", rater data, "even more weight" on YMYL, not all members needed, and Who, How and Why.
• The same page, also cited for the author page and sourcing question, the research self-test, "easily-verified factual errors" and product review evidence.
• Our latest update to the quality rater guidelines: E-A-T gets an extra E for Experience, Google Search Central Blog, Elizabeth Tucker, December 15, 2022. Cited for the date Experience was added and the tax examples.
• Google Says EEAT Is Not A Ranking Factor Nor A Thing That Factors Into Other Factors, Search Engine Roundtable, February 13, 2024. Secondary coverage of Danny Sullivan's February 7, 2024 posts on X, cited for "not a ranking factor" and "doesn't magically make you rank better".
• How Google Fights Disinformation, Google, February 2019. Written before Experience was added; cited for signals that "correlate with trustworthiness and authoritativeness" and PageRank.
• A guide to Google Search ranking systems, Google Search Central (updated December 10, 2025). Cited for the systems listed, helpful content as a retired system, the absence of an E-E-A-T system, and the reviews system.
• Google Search's guidance about AI-generated content, Google Search Central Blog, Danny Sullivan and Chris Nelson, February 8, 2023. Cited for appropriate AI use, "It's just content", bylines, the AI byline line and YMYL reliability.
• Spam policies for Google web search, Google Search Central (updated August 28, 2026). Cited for the definition of scaled content abuse.
• Write high quality reviews, Google Search Central (updated December 10, 2025). Cited for providing evidence of first-hand experience in reviews.
• Optimizing your website for generative AI features on Google Search, Google Search Central (updated July 10, 2026). Cited for "rooted in our core Search ranking and quality systems" and inauthentic mentions.
• AI features and your website, Google Search Central (updated December 10, 2025). Cited for "no additional requirements" for AI Overviews and AI Mode.
• AI Overviews: About last week, Google, Elizabeth Reid, May 30, 2024. Cited for "backed up by top web results", limiting user-generated content, and forums as first-hand sources.
• Bing Webmaster Guidelines, Microsoft Bing, checked September 30, 2026. Cited for the shared foundation, "Authority and trust signals", and sections 11 and 15.
• Publishers and Developers - FAQ, OpenAI Help Center, checked September 30, 2026. Cited for "Any public website can appear in ChatGPT search" and OAI-SearchBot access, and for publishing no trust criteria.
• GEO: Generative Engine Optimization, Aggarwal et al., KDD 2024. A benchmark on a simulated engine, cited for citing sources, quotations and statistics vs an authoritative tone.
• Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied), Ahrefs, December 12, 2025. Vendor data on domains with Domain Rating above 40, cited for mentions vs Domain Rating correlations.
• Update: 38% of AI Overview Citations Pull From The Top 10, Ahrefs, March 2, 2026. Vendor data, cited for the 38% figure and Ahrefs' explanation of the change: the Gemini 3 move and its own improved parsing.
• 76% of AI Overview Citations Pull From the Top 10, Ahrefs, July 21, 2025. Vendor data, cited for the earlier 76% figure.

Argued, not sourced (this guide's reasoning from the sources above):

• The core claim that E-E-A-T is evidence, not a score, as this guide's synthesis of Google's statements.
• The reading of the tension between the help page and Sullivan's post, and that tools' "E-E-A-T scores" are their own estimates.
• The dividing test for the three evidence sources, and where on-site reviews fall.
• That a site's own claims count for little until others back them up, as a reading of 3.3.1's skepticism rule, and that credentials count when others can confirm them.
• Authority being topic by topic, as a reading of the "go-to source for the topic" definition.
• Disclosing conflicts of interest on the page, as an extension of 3.4.
• The chronic condition vs medication illustration, and the recipe vs medication contrast.
• That no engine documents E-E-A-T as a filter or gate for AI citations, and that unsourced percentages circulate, from this guide's review of engine documentation and SERP research in September 2026.
• That, for Google, the same trust evidence carries into AI features, as a reasoned step from its documentation.
• The "no disclosure" row of the authorship table.
• Applying reputation searches to authors' names plus their topic.
• The audit's steps, the data work and judgment call labels, and the priority order.