# E-A-T Audits: How Winston Scores a Site Against Google's Rater Guidelines

**Author:** John Morabito (Founder, /winston)
**Published:** June 13, 2026
**Reading time:** 16 minutes
**Canonical:** https://www.winstondigitalmarketing.com/playbooks/eat-audit-google-rater-guidelines/

An E-A-T audit scores your site against the same Experience, Expertise, Authoritativeness, and Trust signals Google's 168-page Search Quality Rater Guidelines hand to human raters. It is not a ranking dial you turn. It is a quality assessment that asks whether real, qualified people stand behind the content, whether the site is the kind of source Google and the AI engines would cite, and where the trust signals are missing. The output is a scored rubric and a fix list.

## What the rater guidelines actually are

Google employs thousands of external Search Quality Raters. They do not change rankings directly. They rate pages, and Google uses those ratings to train and validate its systems. The instructions they follow are public: a 168-page document Google updates roughly once a year. It is the closest thing to an official spec for "what does Google consider a high-quality, trustworthy page."

The document leans on a few load-bearing ideas. Trust is the center of the model; the guidelines say so explicitly, and the other letters exist to support it. Your-Money-or-Your-Life (YMYL) topics (health, finance, safety, legal, major life decisions) get held to a far higher bar, because bad information there causes real harm. And the newest letter, the extra E for Experience, asks a simple question that catches a lot of AI-generated and ghost-written content: has the author actually done the thing they are writing about?

The one-line version: the rater guidelines are Google telling you, in writing, what a page it wants to reward looks like. An E-A-T audit is the act of grading your own site against that rubric before Google's systems do it for you.

## The rubric we run

We score five dimensions, each on a simple low / medium / high scale with evidence attached, so the result is a defensible grade rather than an opinion.

| Dimension | What we score | The common fail |
|---|---|---|
| Experience (the new E) | First-hand proof: original photos, real test data, "we ran this" language, named operators | Generic content that could have been written by anyone who never touched the subject |
| Expertise | Author credentials matched to topic, depth a non-expert could not fake, accuracy | Anonymous or fabricated author bios, no credentials anywhere on the page |
| Authoritativeness | Is this site a recognized source, who links to and cites it, off-site reputation | A thin site claiming authority it has not earned, zero third-party references |
| Trust | Accurate About page, real contact info, clear ownership, secure and honest, no deceptive design | Stock About paragraph, no physical contact, undisclosed affiliate or sponsored content |
| YMYL handling | For health/finance/legal pages: higher accuracy bar, reviewed-by signals, sourcing | Money or health advice with no qualified reviewer and no citations |

The mechanics are deliberately boring. We pull a representative sample of pages (the money pages plus a slice of the content library), read each one against the dimension, and record a score with a one-line reason and a screenshot. A quality judgment you can show your work on is worth ten times a number you cannot defend. The same evidence-first discipline runs our broader technical SEO audit (https://www.winstondigitalmarketing.com/playbooks/technical-seo-audit-90-minutes-claude/); E-A-T is the trust chapter of that larger pass.

## What fails most often

- **Ghost authors.** Articles attributed to "Admin", a brand name, or a freelancer with no bio, no photo, and no credentials. The guidelines treat unknown authorship on YMYL content as a serious negative, and the AI engines cannot attribute a quote to a person who does not exist.
- **The stock About page.** One paragraph of "we are passionate about quality" with no names, no history, no address. Trust is the center of the model, and the About page is where raters are told to look first.
- **No experience signals.** Content that reads like a competent summary of other content, with nothing first-hand. Exactly what the new E was added to catch, and exactly what most AI-written pages produce by default.
- **Unsourced YMYL claims.** A financial advisor or medical site making confident claims with no citations and no qualified reviewer named. The fastest trust downgrade in the book.
- **Undisclosed commercial intent.** Affiliate or sponsored content presented as neutral. The guidelines have a specific, dim view of deceptive monetization, and it bleeds into the whole domain's trust score.

## From score to fix list

A grade nobody acts on is decoration. Every audit converts into an ordered remediation list, cheapest-and-highest-impact first.

1. **Fix authorship.** Real bylines, real bios, real credentials, real photos, a Person schema block per author connected to the content. Usually the single biggest mover, and it doubles as the entity work in schema markup for AI engines (https://www.winstondigitalmarketing.com/playbooks/schema-markup-for-ai-engines-2026/).
2. **Rebuild the About and contact pages.** Who you are, who runs the company, where you are, how to reach a human. An afternoon of writing, an outsized trust return.
3. **Add experience proof to money pages.** Original photos, data you collected, "here is what we found when we did this" framing. Replace summarized content with lived content.
4. **Source the YMYL pages.** Citations to primary sources, a named reviewer with credentials, dates. Non-negotiable for health, finance, and legal topics.
5. **Disclose everything.** Affiliate relationships, sponsorships, AI assistance where relevant. Honesty reads as trust, both to raters and to readers.

Done well, the same work pays twice. The author credentials, citations, and entity connections that satisfy a quality rater are precisely the signals that get a page cited by ChatGPT and the other AI engines (https://www.winstondigitalmarketing.com/playbooks/how-to-get-cited-by-chatgpt-in-2026/), which increasingly decide who gets named in an answer. Trust is not a separate project from AI visibility. It is the same project.

## Why most agencies skip it

An E-A-T audit implicates the client. It often concludes that the content is anonymous, the About page is a template, and there are no real credentials on the site. That is a harder conversation than handing over a keyword list, and the fix is editorial and organizational, not a meta-tag tweak. Skipping it is easier. It is also why the sites that do the work pull ahead, and why the E-A-T pass is baked into [our SEO service](https://www.winstondigitalmarketing.com/services/seo/) rather than sold as an optional add-on.

## What to do this week

1. Open one of your highest-value pages and ask: who wrote this, and can a reader verify they are qualified? If the answer is no, you have found your first fix.
2. Read your own About page as a skeptic. If it names nobody and proves nothing, rewrite it.
3. Pick your three money pages and add one first-hand experience signal to each.

The free 48-hour audit includes a first-pass E-A-T read on your top pages and the prioritized fix list.

Service: https://www.winstondigitalmarketing.com/services/seo/
Audit: https://www.winstondigitalmarketing.com/contact/#audit