---
title: "E-E-A-T checklist: 42 points to assess a website and brand"
version: "1.0"
date: "2026-10-26"
language: en
guidelines: "Search Quality Rater Guidelines (General Guidelines), version of September 11, 2025"
license: "CC BY 4.0"
author: "Łukasz Rogala"
source: "https://rogala.pro/en/blog/e-e-a-t/"
---

# E-E-A-T checklist: 42 points to assess a website and brand

This checklist helps you check whether a website, and the business behind it, gives readers verifiable evidence that they can be trusted. It has 42 points in seven groups: identity and transparency, experience, expertise, authoritativeness, transaction trust, machine-readable data and negative signals. The questions come from Google's Search Quality Rater Guidelines (version of September 11, 2025), from Google Search documentation and from my own audit practice. Every point says where it comes from, so you can see which requirements are Google's and which are mine.

The checklist is for website owners, editorial teams, SEO specialists and AI agents. It assesses the brand and the website, not a single piece of content. It expands the question list in the E-E-A-T article on rogala.pro.

**The result of this checklist is not a Google rating and does not predict rankings or traffic.** Google says that E-E-A-T itself is not a specific ranking factor, and it does not publish any "E-E-A-T score". The ratings of people who work with the guidelines do not move pages up or down in search results (guidelines p. 6). The result shows where a website gives weak evidence of trustworthiness, or none at all.

## How to use this checklist

**On your own.** Work through steps 1-5 below. For every point, write down the answer and the evidence: the page URL and a short quote, or a description of what you see. Calculate the result with the algorithm in "How to calculate the result".

**With a language model.** Paste this whole file into a chat with any model (ChatGPT, Claude, Gemini, Copilot or another) and add at the end:

```
Assess the website https://your-domain.com against this checklist. Reply in English. If you cannot open the website, ask me to paste the content of the pages you need.
```

A model with web access can open the website itself. If it cannot, it should ask you to paste content: the home page, the about page, the contact page, the footer, the terms, the privacy policy, a few articles and an author page. Points that cannot be checked from the pasted text will be marked "not checked". A model can make mistakes, so check the evidence for every point before you change anything.

**As a skill for an AI agent.** Next to this checklist on the source page there is a skill called `eeat-ocena-strony` in the Agent Skills format, with the English version of this checklist and instructions for the agent.

## Where the points come from

The "Source" column says where the question comes from:

- **Guidelines p. N.** Google's guidelines for search quality raters (General Guidelines), the PDF file in the version of September 11, 2025, 182 pages. N is the page number in that file. In a newer version the page numbers may shift.

- **Google: helpful content.** The Google Search Central documentation page "Creating helpful, reliable, people-first content".

- **Google: spam policies.** The documentation page "Spam policies for Google web search".

- **Google: page experience.** The Google Search documentation page "Understanding page experience in Google Search results".

- **Practice.** My own audit practice. These are not Google requirements, only ways to measure a trust signal. Do not present them as Google guidelines.

I set the pass thresholds myself for every point, including those where the question comes from the guidelines. Quality raters do not add up points; they set a slider from the lowest to the highest rating (guidelines p. 19).

The guidelines describe Trust as the most important part of E-E-A-T, supported by experience, expertise and authoritativeness (pp. 26-27). A page that is untrustworthy for any reason has low E-E-A-T, even if its author is an expert (p. 27). That is why the checklist has three gates that cap the grade regardless of the total score. The gates are my rules, not Google's.

## Step 1. Set the website profile and the assessment mode

Before you start, answer three questions and record the answers in the result:

- **Mode: standard or YMYL?** YMYL topics are topics where content could significantly affect people's health, financial stability or safety, or the welfare of society (guidelines p. 11). The rules for choosing are in the "YMYL mode" section.

- **Sales: yes or no?** Yes when the website sells products or services online or accepts payments.

- **Editorial content: yes or no?** Yes when the website publishes articles, guides, reviews, tests or advice.

These answers decide where "not applicable" is allowed and whether the transaction trust gate applies.

## Step 2. Check whether the brand name also belongs to someone else

This step gives no points, but without it the "Authoritativeness" group can be fiction, because you may end up assessing the reputation of a different company with the same name.

1. Search for the brand name on Google and check who the knowledge panel and the first ten results are about.

2. Ask an AI assistant "What is [brand name]?" and check whether the answer is about the business you are assessing.

3. Check Wikidata and Wikipedia for other entities with the same name.

4. Check whether anything outside the brand's own domain exists that a reputation assessment could rely on.

If there is a name collision (another company, person, place or product), record it in the result before scoring and assess the "Authoritativeness" group only on evidence that can be clearly attributed to the brand you are assessing. Having no sources at all outside the own domain is not a collision; it is a result for the "Authoritativeness" group. If you cannot do this step, record "name collision: not checked".

## Step 3. Collect evidence and answer the 42 points

### How to collect evidence

- **Answers.** Every point gets one of four answers: "yes" (the condition in the "Passes when" column is met), "no", "not applicable" or "not checked".

- **Evidence.** Every "yes" and "no" has evidence: the page URL and a quote of up to 15 words, or a description of what you see and where (for example "the home page footer has no company name"). An answer without evidence is "not checked". Do not invent evidence and do not rely on what you remember about the brand.

- **Not checked.** Use it when you could not check a point: the page did not open, the content was not pasted, or the point needs an account, a purchase or a tool you do not have. Record the reason.

- **Not applicable.** Allowed only for points whose "Passes when" column contains a "Not applicable when..." condition, and only when that condition is met. For all other points "not applicable" is not allowed.

- **Text sample.** Points 1.6, 1.8, 2.1, 2.5, 3.1, 3.2, 3.3, 3.5, 3.6 and 6.3 use one sample: 5 texts (articles, guides, reviews, advice), namely the 3 most important ones from the menu or home page and the 2 most recent ones. In a shop without editorial content, the sample for point 2.1 is 5 product pages. If the website has fewer such pages, take all of them and convert the "3 of 5" or "4 of 5" threshold to 60% or 80% of the number of pages, rounded up.

- **Safety.** Do not enter real personal or payment data, do not pay and do not create accounts just to check a point. Mark a point that requires this as "not checked".

- **Order when time is short.** Start with points 1.1, 1.2, 1.3, 1.6, 2.1, 3.1, 3.4 (in YMYL mode), 4.1, 4.2, 4.4, 5.1, 5.2, 6.1, 6.2, 7.1 and 7.3. Mark the points you do not have time for as "not checked" and calculate the result the same way as in a full assessment. Unchecked points in group 7 trigger the negative signals gate, so such a review gives a grade of C at best.

The "Weight" column gives the weight of a single point in standard mode and in YMYL mode, in points out of 100, when all points of the group are applicable. Weights follow from the group share (see "How to calculate the result"), and with "not applicable" answers they change according to the algorithm. The weights in the tables are rounded to two decimal places, so their sum may differ from 100 by 0.01; calculate the result from the group shares, not from the rounded weights.

### 1. Identity and transparency (8 points, gate)

The guidelines ask raters to find out who is responsible for the website and who created the content (p. 16). Pages with payment functionality that have an unsatisfying amount of contact and customer service information, and YMYL pages that say too little about who is responsible, should get a Low rating (p. 63). YMYL pages or pages that handle sensitive data with no information at all about the website or the content creator should get the Lowest rating (p. 36).

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 1.1 | Does the "About us" or "About me" page describe the entity with facts? | Open the page about the business or author from the menu or footer (e.g. /about). | The page says who runs the website, since when and what they do, with facts (dates, places, names, numbers), not just advertising slogans. | 2.5 / 2.75 | Guidelines pp. 16, 18 and 27 |
| 1.2 | Does the website give full details of the entity responsible for it? | Footer, contact page, terms, privacy policy. | The legal name, address and a company registration or tax number (for example a Companies House number in the UK, or NIP, KRS or CEIDG in Poland) are at most two clicks from the home page. Not applicable when it is a personal website with no sales and no YMYL content. | 2.5 / 2.75 | Guidelines pp. 16 and 35; registration number: practice |
| 1.3 | Is there a working way to contact a human, appropriate to the website? | Contact page; if you can, send a short question. | There is at least one way to contact a human (email, phone, address). Websites with sales and YMYL websites offer more than a contact form alone. | 2.5 / 2.75 | Guidelines pp. 18 and 63 |
| 1.4 | Do the entity's details match the public registry? | Compare the name, address and numbers with the business registry (for example Companies House, the national business register, VIES for EU VAT numbers). | The name, address and numbers match the registry. Not applicable when it is a personal website with no sales and no YMYL content. | 2.5 / 2.75 | Practice |
| 1.5 | Is the website owned by the entity the website names as its owner? | Compare the footer, terms, privacy policy, structured data and the public domain registry (WHOIS). | The sources do not contradict each other. Hidden data in the domain registry is not a contradiction. | 2.5 / 2.75 | Practice |
| 1.6 | Are texts with advice and opinions signed with a first and last name? | Text sample. | Every text in the sample has an author named with a first and last name; a byline such as "admin", "editorial team" or no byline at all is not enough. Not applicable when the website does not publish editorial content. | 2.5 / 2.75 | Guidelines p. 16; Google: helpful content |
| 1.7 | Does the website say how its content is created? | Look for an editorial policy, a methodology or a process description (footer, about page). | The description says who creates the content, how it is checked and how errors are corrected, and if content is substantially generated automatically or with AI, it says so too. Not applicable when the website does not publish editorial content (not allowed in YMYL mode). | 2.5 / 2.75 | Google: helpful content (how the content was created); editorial policy: practice |
| 1.8 | Are commercial relationships disclosed? | Text sample: look for affiliate links, product rankings, sponsored articles and how they are labelled. | Every text in the sample that earns money through recommendations has a label next to the content that readers can understand (e.g. "affiliate link", "sponsored content"). Not applicable when the sample has no such texts. | 2.5 / 2.75 | Guidelines p. 27 (conflict of interest) |

### 2. Experience (5 points)

The "E" for Experience: whether the content creator has personally used the product or lived through what they write about, when the topic requires it (guidelines p. 26). A restaurant review written by someone who has never eaten at the restaurant is the guidelines' example of a lack of experience (p. 60).

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 2.1 | Does the content show first-hand evidence? | Text sample (in a shop without editorial content: 5 product pages). | At least 3 of 5 pages have own photos, measurements, screenshots or a description of actual use, rather than stock photos and paraphrases of other people's texts. | 2.4 / 1.6 | Guidelines p. 26; Google: helpful content |
| 2.2 | Are the key photos original? | Run a reverse image search (e.g. Google Lens) on 3 key photos: the team, products, completed projects or photos in reviews. | None of the 3 photos comes from a stock photo library or from another website. Not applicable when the website has no such photos. | 2.4 / 1.6 | Practice |
| 2.3 | Do tests, comparisons and rankings describe their method? | Texts such as tests, comparisons, rankings, "best of" lists. | The text says what was tested, how many products, how and when, and shows the results with evidence (photos, measurements) in enough detail to repeat it. Not applicable when the website publishes no tests, comparisons or rankings. | 2.4 / 1.6 | Google: helpful content (how the content was created); guidelines p. 37 (a false claim of own testing is deception) |
| 2.4 | Is the business's hands-on experience visible? | Completed projects, case studies, portfolio, photos of the premises, history of the business or the publication. | There are specifics that can be verified (projects, dates, numbers, places), not just "years of experience". | 2.4 / 1.6 | Practice |
| 2.5 | Does the content add something others do not have? | Take 2 texts from the sample (in a shop without editorial content: 2 product descriptions), compare them with the top ten Google results for the same query and search for 2-3 sentences from each in quotation marks to rule out copying. | Both texts have their own data, examples, analysis or experience that the compared results do not have, and neither is a copy. | 2.4 / 1.6 | Google: helpful content; guidelines p. 21 (originality) and pp. 43-44 (how to identify copied content) |

### 3. Expertise (6 points)

The "E" for Expertise: whether the content creator has the knowledge or skill the topic requires (guidelines p. 26). The assessment should rest on verifiable credentials, not on a claim of "I'm an expert" (p. 63). On YMYL topics, content in which people share their own experience can have high E-E-A-T, but some information and advice must come from experts (p. 28).

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 3.1 | Do authors have profile pages? | Click the author byline on the texts in the sample. | The byline leads to an author page at its own URL (on a one-person website: to the page about the author) that describes the author's experience and the topics they write about, and is not excluded from indexing (noindex). Not applicable when the website does not publish editorial content. | 2.5 / 3.67 | Google: helpful content (bylines lead to information about the author); URL and indexing: practice |
| 3.2 | Can the authors' credentials be confirmed outside the website? | Search for every author in the sample: professional profile, publications, talks, professional register. | Every author in the sample has at least one independent confirmation of their credentials outside the website being assessed. Not applicable when the website does not publish editorial content. | 2.5 / 3.67 | Guidelines pp. 25 and 63 |
| 3.3 | Do the authors' credentials match their topics? | Match each author with the topic of their text in the sample. | No author in the sample signs texts in fields where they show no knowledge or experience (e.g. the same person writing about loans, medicines and home renovation). Not applicable when the website does not publish editorial content. | 2.5 / 3.67 | Guidelines pp. 22 and 26 |
| 3.4 | Is YMYL advice written or reviewed by a qualified specialist? | Bylines and review notes on texts with YMYL advice. | Every checked text with YMYL advice has an author or reviewer with verifiable qualifications (e.g. a licence number, professional accreditation). Not applicable in standard mode. | 2.5 / 3.67 | Guidelines pp. 28 and 36; Google: helpful content |
| 3.5 | Are publication and update dates genuine? | Dates on the texts in the sample; if you can, compare the text with a copy in the Wayback Machine. | The texts in the sample show a publication or update date, and you found no update date changed without a change to the content. Not applicable when the website does not publish editorial content. | 2.5 / 3.67 | Google: helpful content (changing dates without changing content) |
| 3.6 | Do factual claims have sources? | Text sample: numbers, data, legal and medical claims. | In at least 4 of 5 texts, numerical and legal claims have a source that can be opened. Not applicable when the website does not publish editorial content. | 2.5 / 3.67 | Google: helpful content; guidelines pp. 64 and 76 (examples) |

### 4. Authoritativeness (7 points)

Reputation outside the website itself. The guidelines tell raters to be sceptical of what a website says about itself and to look for independent reviews, articles, recommendations by experts and Wikipedia articles (pp. 22-23). Reviews on the website itself are not independent evidence. A small local business may have little information online, and according to the guidelines this indicates neither high nor low quality (p. 25). In that case the points in this group will come out weak, so say so in the result: the checklist measures the amount of evidence, not how good the business is.

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 4.1 | Do independent, credible sources write about the brand? | Search for `brand name -site:example.com` and `"example.com" -site:example.com`. | There are editorial mentions (articles, rankings by independent organisations, Wikipedia), not only directories, press releases and profiles run by the brand. | 2.86 / 2.57 | Guidelines pp. 22-23 |
| 4.2 | Do reviews outside the website exist, and are they not mostly negative? | Search for `brand name reviews -site:example.com`; the Google Business Profile, review websites. Read the reviews, not just the star ratings. | There are reviews outside the website, positive ones prevail and there are no repeated credible reports of fraud, undelivered orders or unrefunded money. | 2.86 / 2.57 | Guidelines pp. 23 and 36 |
| 4.3 | Does the business respond to negative reviews? | The same websites as in 4.2. | Most negative reviews from the last 12 months have a factual response from the business. Not applicable when there are no negative reviews outside the website. | 2.86 / 2.57 | Practice |
| 4.4 | Does the brand exist in knowledge bases? | Wikidata, Wikipedia, the Google knowledge panel for the brand name. | There is a Wikidata or Wikipedia entry, or a knowledge panel, with correct data about the brand being assessed. | 2.86 / 2.57 | Practice |
| 4.5 | Does the link profile look earned rather than bought? | A link analysis tool (e.g. Ahrefs, Majestic, Semrush): referring domains, anchor texts, growth rate. | Exact-match commercial anchor texts do not dominate and there are no sudden spikes in links from low-quality websites. | 2.86 / 2.57 | Practice |
| 4.6 | Is the brand cited in its industry? | Search for the brand name together with the industry: trade media, reports, conferences. | At least 3 independent citations or talks in the last 12 months. | 2.86 / 2.57 | Practice |
| 4.7 | Do people search for the brand by name? | Google Search Console (queries with the brand name, if you have access), Google Trends or a keyword research tool. | Searches for the brand name are above zero and have not been declining over the last 12 months. | 2.86 / 2.57 | Practice (measures recognition, not reputation; the guidelines tell raters to ignore traffic data when judging reputation, p. 22) |

### 5. Transaction trust (7 points, gate for websites with sales and YMYL websites)

The guidelines say that contact and customer service information is extremely important for websites that handle money, and for shops they ask raters to check the policies on payment, exchanges and returns (p. 18). Online stores need secure payment systems and reliable customer service (p. 26).

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 5.1 | Does the whole website run on HTTPS without mixed content? | The address bar and the browser console on the home page, a product or service page and a page with a form. | All checked pages load over HTTPS, with no mixed content warnings. | 2.57 / 2.57 | Google: page experience; guidelines p. 26 |
| 5.2 | Are the returns, complaints and delivery policies easy to find? | Navigate from a product page or a paid service page. | The policies are at most two clicks from the product page. Not applicable when the website has no sales. | 2.57 / 2.57 | Guidelines p. 18 |
| 5.3 | Are payment methods and their security described? | Cart, payment information page, terms. | Payment methods are described before the website asks for personal or payment data. Not applicable when the website has no sales. | 2.57 / 2.57 | Guidelines pp. 18 and 26 |
| 5.4 | Is the full cost known before data has to be entered? | Go through the checkout up to the payment step, without paying and without entering real data. | The final cost, including delivery and fees, is visible before personal data is entered. Not applicable when the website has no sales. | 2.57 / 2.57 | Practice |
| 5.5 | Is personal data collected with information about its purpose? | Forms (contact, newsletter, account) and the privacy policy. | Every form has information about the purpose of data collection or a link to a privacy policy that describes this purpose and gives a contact for data matters. Not applicable when the website has no forms and no user accounts. | 2.57 / 2.57 | Practice; guidelines p. 36 (a website that handles personal data needs extensive contact information) |
| 5.6 | Does customer service actually respond? | Send a question through the contact channel and measure the response time. | A reply from a human arrives within 2 business days. Not applicable when the website has no sales. | 2.57 / 2.57 | Guidelines p. 26 |
| 5.7 | Is the website free of deceptive patterns? | Checkout, pop-ups, consent banner, subscription cancellation. | There is no forced consent, no fake countdown timers or stock counters, no hidden subscriptions, no buttons that do something other than what they show and no pop-ups that cannot be closed without an action that benefits the website. | 2.57 / 2.57 | Guidelines pp. 37-38 (deceptive buttons and pop-ups) |

### 6. Machine-readable data (5 points)

This whole group is my practice. A rater reads the page, not the code, and I know of no evidence that structured data by itself raises E-E-A-T. This group checks whether the evidence from groups 1-5 is recorded in a way that search engines and language models can read and connect to the right business and people.

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 6.1 | Does structured data describe the entity responsible for the website? | The Rich Results Test or the schema.org validator on the home page. | There is `Organization` or `LocalBusiness` (on a personal website `Person`) with a name, an address or city, contact details and a logo (a photo for `Person`). | 1.6 / 1.0 | Practice |
| 6.2 | Does `sameAs` connect the brand with its profiles? | The home page code (JSON-LD). | `sameAs` points to profiles that really belong to the brand, and to the Wikidata entry if one exists. | 1.6 / 1.0 | Practice |
| 6.3 | Are authors described in structured data? | The code of the author page and of an article from the sample. | The author is described as a `Person` with a first and last name and topics (`knowsAbout`), the article points to them in the `author` field, and the data matches the page content. Not applicable when the website does not publish editorial content. | 1.6 / 1.0 | Practice |
| 6.4 | Are the name, address and phone number the same everywhere? | Compare the website, the Google Business Profile and 3 business directories. | The name, address and phone number match in all checked sources. Not applicable when the business has no Google Business Profile and no business directory listings. | 1.6 / 1.0 | Practice |
| 6.5 | Can the crawlers that should see the website open it? | The robots.txt file; the server response for different user agents (e.g. Googlebot, Bingbot, AI assistant crawlers), for example `curl -A "<user-agent>" <url>`; if you have access, the URL Inspection tool in Google Search Console. | Googlebot and Bingbot get the page (status 200, the same content as in a browser), and blocks on other crawlers are set in robots.txt rather than caused by server errors or firewall challenge screens. A firewall may check the IP address of the real Googlebot: if the `curl` test is not conclusive, answer "not checked". | 1.6 / 1.0 | Practice |

### 7. Negative signals (4 points, reversed scoring, gate)

In this group "yes" means the problem is absent. Points 7.1, 7.2 and 7.4 match reasons for the Lowest rating in the guidelines, and point 7.3 matches Google's spam policies. In January 2025 Google aligned the guidelines' Lowest and Low sections with these policies (p. 182). This group has no "not applicable" answer.

| No. | Question | How to check | Passes when | Weight | Source |
|---|---|---|---|---|---|
| 7.1 | Is the website free of mass-produced content without value? | Look at several pages of the same type: identical templates with a swapped keyword, serial publication dates, leftover AI model language (e.g. "As an AI language model"). | There are no such patterns, and pages built from one template (e.g. product pages, city pages) have content that the other pages in the series do not have. | 1.75 / 1.75 | Guidelines pp. 42 and 44; Google: spam policies (scaled content abuse) |
| 7.2 | Does the domain avoid hosting third-party content placed there to benefit from its rankings? | Look at directories and subdomains that do not fit the website's topic (e.g. casinos, payday loans, coupons on a website about something else). | There are no sections with content from other entities published mainly to benefit from the domain's position in search results. | 1.75 / 1.75 | Guidelines p. 41; Google: spam policies (site reputation) |
| 7.3 | Does the website avoid hiding content and avoid showing crawlers something different from people? | Compare the source code with what the browser shows; look for text in the background colour or off-screen, and for different content served to the Googlebot user agent. | There is no hidden text and no cloaking. | 1.75 / 1.75 | Google: spam policies (cloaking, hidden text) |
| 7.4 | Does the website avoid misleading people about the business, authors, prices or status? | Compare the business details with the registry, the product page price with the price in the cart, check the authors and their credentials outside the website, the address of a physical store, claimed certificates and awards. | There are no contradictions, no made-up author profiles and no false claims. | 1.75 / 1.75 | Guidelines pp. 36-37 and 63; Google: helpful content (fabricated creator profiles are deception) |

## Step 4. How to calculate the result

### Group shares

| Group | Points | Share, standard mode | Share, YMYL mode |
|---|---|---|---|
| 1. Identity and transparency | 8 | 20 | 22 |
| 2. Experience | 5 | 12 | 8 |
| 3. Expertise | 6 | 15 | 22 |
| 4. Authoritativeness | 7 | 20 | 18 |
| 5. Transaction trust | 7 | 18 | 18 |
| 6. Machine-readable data | 5 | 8 | 5 |
| 7. Negative signals | 4 | 7 | 7 |
| Total | 42 | 100 | 100 |

### Algorithm

1. **Applicable points.** In each group, count the applicable points S, that is all points of the group except those answered "not applicable".

2. **Passed points.** Count the "yes" answers Y. Answers "no" and "not checked" score 0. There are no partial answers.

3. **Group result.** The group result is Y / S, and the group contribution is the group's share in the chosen mode times Y / S. The weight of one point is the group share divided by S, so a "not applicable" answer spreads that point's weight over the remaining points of the same group.

4. **Overall result.** Result = 100 × sum of contributions / sum of the shares of groups where S is greater than zero. When every group has applicable points, the shares add up to 100 and the result is simply the sum of contributions. A group in which all points are "not applicable" drops out of the calculation.

5. **Rounding.** Calculate with full numbers and round the overall result to one decimal place only at the end.

6. **Letter grade.** Set the grade from the rounded result using the scale, then apply the gates.

7. **Uncertain group.** If more than half of a group's applicable points are "not checked", mark the group as uncertain in the result and do not present its percentage as a strength or a weakness. This does not change the numbers.

### Scale

| Grade | Result |
|---|---|
| A | from 90.0 to 100 |
| B | from 75.0 to 89.9 |
| C | from 60.0 to 74.9 |
| D | from 45.0 to 59.9 |
| E | from 30.0 to 44.9 |
| F | from 0 to 29.9 |

### Gates

A gate caps the letter grade but does not change the numerical result. When several gates apply, the lowest cap wins. The final grade is the worse of two: the letter from the scale or the cap from a gate.

- **Identity gate.** A group 1 result below 50% means a grade of D at best. In that case it is unclear who is responsible for the website.

- **Transaction trust gate.** Applies when the website has sales or is assessed in YMYL mode: a group 5 result below 50% means a grade of D at best.

- **Negative signals gate.** Any group 7 point answered "no" or "not checked" means a grade of C at best. The absence of a negative signal has to be checked for the point to pass, so an unchecked point triggers the gate just like a failed one.

The group result is calculated exactly as in the algorithm, so "not checked" lowers it for the gates as well. Exactly 50% passes a gate.

### Worked example

An online shop with a blog, standard mode (the product range is not YMYL), sales: yes, editorial content: yes.

| Group | Answers | Passed / applicable | Group result | Share | Contribution |
|---|---|---|---|---|---|
| 1. Identity | 1.7 no, all others yes | 7 / 8 | 87.5% | 20 | 17.50 |
| 2. Experience | 2.3 not applicable (no tests), 2.2 no, all others yes | 3 / 4 | 75.0% | 12 | 9.00 |
| 3. Expertise | 3.4 not applicable (standard mode), 3.2 no, all others yes | 4 / 5 | 80.0% | 15 | 12.00 |
| 4. Authoritativeness | 4.4 no, 4.7 not checked, all others yes | 5 / 7 | 71.4% | 20 | 14.29 |
| 5. Transaction trust | 5.6 not checked, all others yes | 6 / 7 | 85.7% | 18 | 15.43 |
| 6. Machine-readable data | 6.3 no, 6.4 no, all others yes | 3 / 5 | 60.0% | 8 | 4.80 |
| 7. Negative signals | 7.3 not checked, all others yes | 3 / 4 | 75.0% | 7 | 5.25 |
| Total | | 31 / 40 | | 100 | 78.27 |

Result: 78.3 out of 100 (the sum of contributions calculated with full numbers is 78.264). The scale gives a B. Identity gate: 87.5%, passes. Transaction trust gate: 85.7%, passes. Negative signals gate: point 7.3 was not checked, so the grade is C at best. Final: 78.3 out of 100, grade C, 3 of 40 applicable points not checked, 2 points not applicable. If point 7.3 had been checked with a "yes", group 7 would contribute 7.00, the result would be 80.0 and the grade B.

An example of a group dropping out of the calculation: if, with all other answers unchanged, every point in group 3 were "not applicable", the result = 100 × (78.264 - 12.00) / (100 - 15) = 78.0.

## YMYL mode

### When to choose YMYL mode

The guidelines divide YMYL topics into four groups: health or safety, financial security, government, civics and society, and other topics that could hurt people or the welfare of society (p. 11). YMYL is a spectrum, not a list of industries: the guidelines show topics that are clearly YMYL, topics that may be YMYL and topics that are not (p. 12). Rules for choosing the mode:

- Choose YMYL mode when the website or the assessed section publishes advice or information on YMYL topics, or sells products and services where a mistake could seriously harm someone (e.g. prescription drugs, loans, insurance, legal services).

- When you are not sure, ask the question from the guidelines: could a small mistake in this content seriously harm someone (pp. 11-12)? If the answer is "yes", use YMYL mode.

- One assessment has one mode. If only part of the website covers YMYL topics and you are assessing the whole website, choose YMYL mode.

- Record a one-sentence justification of the mode in the result.

### What YMYL mode changes

- **Group shares.** You calculate with the "Share, YMYL mode" column: expertise and identity weigh more, experience and machine-readable data weigh less.

- **Points without "not applicable".** Points 1.7 (how content is created) and 3.4 (a specialist for YMYL advice) must be answered "yes", "no" or "not checked".

- **Reputation according to experts.** In group 4, judge the reputation of a YMYL website by what experts in the field say; recommendations from professional societies are strong evidence (p. 22).

- **Experience versus expertise.** Content in which people share their own experience can have high E-E-A-T if it is safe and consistent with expert consensus, but information and advice such as treatment or investment options should come from experts (p. 28).

### Gates in YMYL mode

- **Identity gate.** A group 1 result below 50% means a grade of D at best.

- **Transaction trust gate.** In YMYL mode it always applies, even without sales: a group 5 result below 50% means a grade of D at best.

- **Negative signals gate.** Any group 7 point answered "no" or "not checked" means a grade of C at best.

## Step 5. Record the result

Record the result in this layout. A language model replies in the user's language, and the point numbers stay the same.

```
E-E-A-T assessment: <domain>, <assessment date>
Mode: standard | YMYL (<one-sentence justification>)
Profile: sales yes | no; editorial content yes | no
Name collision: none | <description> | not checked
Result: <number with one decimal place> out of 100, grade <letter>
Gates: none triggered | <which gate and why>
Not checked: <number> of <number of applicable points>; uncertain groups: none | <list>
Not applicable: <number> (<point numbers and reason>)
```

Group table:

| Group | Passed / applicable | Group result | Share | Contribution |
|---|---|---|---|---|
| 1. Identity and transparency | | | | |

Point table, all 42 rows:

| Point | Answer | Evidence (URL, quote of up to 15 words, or description) |
|---|---|---|
| 1.1 | yes / no / not applicable / not checked | |

For "not applicable", give the condition that is met, and for "not checked", the reason.

What to fix first, at most 5 points: first the points that triggered a gate, then the points answered "no" in order of highest weight in the chosen mode, and finally the unchecked points worth checking. One sentence per point: what to do.

Always end the result with this sentence: "This result is not a Google rating. Google does not publish any E-E-A-T score, and this assessment does not predict search rankings or traffic."

## What this checklist does not do

- It does not predict rankings or traffic and does not say how Google rates the website.

- It does not assess the quality of a single text. For that, use Google's content self-assessment questions in the helpful content documentation.

- It does not replace a legal audit (data protection, consumer law): it checks whether information is present and consistent, not whether it complies with the law.

- It does not reduce reputation to one number: the points in group 4 require reading reviews and articles, not counting stars.

## Version and licence

- Version: 1.0

- Date: 2026-10-26

- Guidelines version: General Guidelines of September 11, 2025 (https://guidelines.raterhub.com/searchqualityevaluatorguidelines.pdf)

- Google Search documentation: https://developers.google.com/search/docs/fundamentals/creating-helpful-content, https://developers.google.com/search/docs/essentials/spam-policies, https://developers.google.com/search/docs/appearance/page-experience

- Author: Łukasz Rogala

- Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, adapt and share this checklist, including commercially, with attribution: "Łukasz Rogala, rogala.pro", and a link to the source.

- Source: https://rogala.pro/en/blog/e-e-a-t/