Home
cd ../playbooks
Marketing & ContentIntermediate

GEO Content Quality: Score Your Pages on E-E-A-T

Grade any page on Experience, Expertise, Authoritativeness, and Trustworthiness, then get the specific rewrites that make AI engines trust it enough to cite

10 minutes
By Zubair TrabzadaSource
#geo#content#ai-search#copywriting#seo

Google's December 2025 rater guidelines extended E-E-A-T from health and finance pages to every competitive query. Your content can be accurate and well-written and still get skipped by ChatGPT because there is no author byline, no first-hand testing, and no date on the page.

Who it's for: content marketers, SEO consultants, agency strategists, technical writers, in-house content teams, founders publishing their own blog

Example

"Score our blog and pricing page on E-E-A-T" → A GEO-CONTENT-ANALYSIS.md with a 0-100 score broken into Experience, Expertise, Authoritativeness and Trustworthiness out of 25 each, a per-page table of word count and heading structure, flagged passages with rewrite suggestions, and a topical authority modifier from +10 to -5

CLAUDE.md Template

New here? 3-minute setup guide → | Already set up? Copy the template below.

# GEO Content Quality & E-E-A-T Assessment

AI search platforms do not just find content — they evaluate whether content deserves to be cited. The primary framework for that evaluation is **E-E-A-T** (Experience, Expertise, Authoritativeness, Trustworthiness), which per Google's December 2025 Quality Rater Guidelines update now applies to **all competitive queries**, not only YMYL (Your Money Your Life) topics. Content scoring high on E-E-A-T is far more likely to be cited by ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews.

Evaluate content through two lenses:

1. **E-E-A-T signals** — does the content demonstrate real expertise and trust?
2. **AI citability** — is the content structured so AI platforms can extract and cite specific claims?

## How to Run an Assessment

1. Fetch the target pages — homepage, key blog posts, service and product pages.
2. Score E-E-A-T across the four dimensions (25 points each).
3. Assess content quality metrics: structure, readability, depth.
4. Check for low-quality AI content signals.
5. Evaluate topical authority across the site.
6. Score and write `GEO-CONTENT-ANALYSIS.md`.

---

## E-E-A-T Framework (100 points total)

### Experience — 25 points

First-hand knowledge and direct involvement with the topic. AI platforms increasingly distinguish between content that reports on a topic and content from someone who has done it.

| Signal | Points | How to Score |
|---|---|---|
| First-person accounts ("I tested...", "We implemented...") | 5 | 5 if present and specific, 3 if generic, 0 if absent |
| Original research or data not available elsewhere | 5 | 5 if original data, 3 if references original work, 0 if none |
| Case studies with specific results | 4 | 4 if detailed with numbers, 2 if general, 0 if none |
| Screenshots, photos, or evidence of direct use | 3 | 3 if authentic evidence, 1 if stock/generic, 0 if none |
| Specific examples from personal experience | 4 | 4 if specific and unique, 2 if somewhat specific, 0 if generic |
| Demonstrations of process (not just outcome) | 4 | 4 if step-by-step from experience, 2 if partial, 0 if none |

**Flag as weak Experience:**

- Content that summarizes what other sources say without adding a new perspective
- Generic advice that could apply to any situation ("It depends on your needs")
- No mention of actual usage, testing, or direct involvement
- Hedging language suggesting no direct knowledge ("reportedly", "supposedly", "some say")

### Expertise — 25 points

Demonstrated knowledge depth and professional competence in the subject matter.

| Signal | Points | How to Score |
|---|---|---|
| Author credentials visible (bio, degrees, certifications) | 5 | 5 if full credentials, 3 if basic bio, 0 if no author |
| Technical depth appropriate to topic | 5 | 5 if thorough technical treatment, 3 if adequate, 0 if superficial |
| Methodology explanation (how conclusions were reached) | 4 | 4 if clear methodology, 2 if some explanation, 0 if none |
| Data-backed claims (statistics, research citations) | 4 | 4 if well-sourced, 2 if some data, 0 if unsupported claims |
| Industry-specific terminology used correctly | 3 | 3 if accurate specialized language, 1 if basic, 0 if errors |
| Author page with detailed professional background | 4 | 4 if dedicated author page, 2 if brief bio, 0 if none |

**Flag as weak Expertise:**

- Claims without supporting evidence or sources
- Surface-level coverage of complex topics
- Misuse of technical terminology
- No visible author, or an author without relevant credentials
- Content that is broad and generic rather than deep and specific

### Authoritativeness — 25 points

Recognition by others as a credible source on the topic.

| Signal | Points | How to Score |
|---|---|---|
| Inbound citations from authoritative sources | 5 | 5 if cited by major sources, 3 if some citations, 0 if none |
| Author quoted or cited in press/media | 4 | 4 if media mentions, 2 if industry mentions, 0 if none |
| Industry awards or recognition mentioned | 3 | 3 if relevant awards, 1 if tangential, 0 if none |
| Speaker credentials (conferences, events) | 3 | 3 if listed, 0 if none |
| Published in peer-reviewed or respected outlets | 4 | 4 if tier-1 publications, 2 if industry outlets, 0 if none |
| Comprehensive topic coverage (topical authority) | 3 | 3 if site covers topic thoroughly, 1 if some coverage, 0 if isolated |
| Brand mentioned on Wikipedia or authoritative references | 3 | 3 if Wikipedia, 2 if other encyclopedic refs, 0 if none |

**Flag as weak Authoritativeness:**

- Single-topic site with no depth of coverage
- No external validation of expertise claims
- No backlinks from authoritative sources
- Claims of authority without evidence (self-proclaimed "expert")

### Trustworthiness — 25 points

Signals that the content and its publisher are reliable and transparent.

| Signal | Points | How to Score |
|---|---|---|
| Contact information visible (address, phone, email) | 4 | 4 if full contact info, 2 if email only, 0 if none |
| Privacy policy present and linked | 2 | 2 if present, 0 if absent |
| Terms of service present | 1 | 1 if present, 0 if absent |
| HTTPS with valid certificate | 2 | 2 if valid HTTPS, 0 if not |
| Editorial standards or corrections policy | 3 | 3 if documented, 1 if implicit, 0 if none |
| Transparent about business model and conflicts | 3 | 3 if clear disclosures, 1 if some, 0 if none |
| Reviews and testimonials from real customers | 3 | 3 if verified reviews, 1 if testimonials, 0 if none |
| Accurate claims (no misinformation detected) | 4 | 4 if all claims accurate, 2 if mostly accurate, 0 if errors found |
| Clear affiliate/sponsorship disclosures | 3 | 3 if properly disclosed, 0 if undisclosed or absent |

**Flag as weak Trustworthiness:**

- No contact information or physical address
- Missing privacy policy or terms
- Undisclosed affiliate links or sponsored content
- Claims that are verifiably false or misleading
- No way to contact the publisher for corrections

---

## Content Quality Metrics

### Word Count Benchmarks

These are **floors, not targets**. More words does not mean better content. Each figure is the minimum length to cover a topic adequately for AI citability.

| Page Type | Minimum Words | Ideal Range | Notes |
|---|---|---|---|
| Homepage | 500 | 500-1,500 | Clear value proposition, not a wall of text |
| Blog post | 1,500 | 1,500-3,000 | Thorough but focused |
| Pillar content / ultimate guide | 2,000 | 2,500-5,000 | Comprehensive topic coverage |
| Product page | 300 | 500-1,500 | Descriptions, specs, use cases |
| Service page | 500 | 800-2,000 | What, how, why, for whom |
| About page | 300 | 500-1,000 | Company or person story and credentials |
| FAQ page | 500 | 1,000-2,500 | Thorough answers, not one-liners |

### Readability

- **Target Flesch Reading Ease**: 60-70 (8th-9th grade level).
- Not a direct ranking factor, but it affects citability. AI platforms prefer clear, unambiguous content.
- Overly academic writing (score below 30) reduces citability for general queries.
- Overly simple writing (score above 80) may lack the depth needed for expertise signals.

Estimate without a tool:

- Average sentence length: 15-20 words
- Average paragraph length: 2-4 sentences
- Jargon: define on first use
- Passive voice: under 15% of sentences

### Paragraph Structure for AI Parsing

AI platforms extract content at the paragraph level. Each paragraph should be a self-contained unit of meaning.

- **2-4 sentences** per paragraph. One-sentence paragraphs are weak; 5+ sentences are hard to extract.
- **One idea per paragraph.** Do not mix topics inside a paragraph.
- **Lead with the key claim.** The first sentence carries the main point.
- **Support with evidence.** Remaining sentences supply data, examples, or context.
- **Quotable standalone.** Each paragraph should make sense extracted in isolation.

### Heading Structure

- One H1 per page: the primary topic.
- H2 for major sections, each a distinct subtopic.
- H3 for subsections nested under the relevant H2.
- No skipped levels. Never jump H1 to H3.
- Descriptive headings: "How to Optimize for AI Search", not "Section 2".
- Question-based headings where appropriate. They map directly to AI queries.

### Internal Linking

- Every content page should link to 3-5 related pages on the same site.
- Use descriptive anchor text, never "click here".
- Build topic clusters: a pillar page linked to and from all related subtopic pages.
- Orphan pages, with no internal links pointing to them, are rarely cited by AI.

---

## AI Content Assessment

AI-generated content is acceptable per Google's March 2024 clarification, as long as it demonstrates genuine E-E-A-T signals and has human oversight. The question is not how content was created but whether it provides value.

### Signs of Low-Quality AI Content (flag these)

| Signal | Description |
|---|---|
| Generic phrasing | "In today's fast-paced world...", "It's important to note that...", "At the end of the day..." |
| No original insight | Content that only rephrases widely available information |
| Lack of first-hand experience | No personal anecdotes, case studies, or specific examples |
| Perfect but empty structure | Well-formatted headings with shallow content beneath them |
| No specific examples | Abstract explanations without concrete instances |
| Repetitive conclusions | Each section ends with a variation of the same point |
| Hedging overload | "Generally speaking", "In most cases", "It depends on various factors" without naming the factors |
| Missing human voice | No opinions, preferences, or professional judgment expressed |
| Filler content | Paragraphs that could be deleted without losing information |
| No data or sources | Claims presented as facts without attribution or evidence |

### High-Quality Content Signals (regardless of how it was produced)

| Signal | Description |
|---|---|
| Original data | Surveys, experiments, benchmarks, proprietary analysis |
| Specific examples | Named products, companies, dates, numbers |
| Contrarian or nuanced views | Disagreement with conventional wisdom, backed by reasoning |
| First-person experience | "When I tested this..." or "Our team found..." |
| Updated information | References to recent events and current data |
| Expert opinion | Clear professional judgment, not just facts |
| Practical recommendations | Specific, actionable advice, not vague guidance |
| Trade-offs acknowledged | "This approach works well for X but not for Y because..." |

---

## Content Freshness

Check for visible `datePublished` and `dateModified` in both the content and the structured data. Content without dates is treated as less trustworthy by AI platforms. Dates should be specific ("January 15, 2026"), not vague ("recently").

| Criterion | Score |
|---|---|
| Updated within 3 months | Excellent — current and relevant |
| Updated within 6 months | Good — still reasonably current |
| Updated within 12 months | Acceptable — may need refresh |
| Updated 12-24 months ago | Warning — review for accuracy |
| No date, or 24+ months old | Critical — AI platforms may deprioritize |

Some content stays relevant regardless of age. Mark content evergreen if it covers fundamental concepts that do not change (physics, basic math, legal definitions), is labeled as a lasting reference, and contains no time-dependent claims ("the latest", "currently", "in 2024").

---

## Topical Authority

Topical authority measures whether a site covers a topic comprehensively rather than superficially. AI platforms prefer citing recognized authorities.

Assess it on five questions:

1. **Content breadth** — does the site have multiple pages covering different aspects of its core topic?
2. **Content depth** — do individual pages go deep into subtopics?
3. **Topic clustering** — are pages organized into logical groups with internal linking?
4. **Content gaps** — are there obvious subtopics the site should cover but does not?
5. **Competitor comparison** — do competitors cover subtopics this site misses?

| Level | Description | Score Impact |
|---|---|---|
| Authority | 20+ pages covering the topic comprehensively, strong clustering | +10 bonus |
| Developing | 10-20 pages with some clustering | +5 bonus |
| Emerging | 5-10 pages on topic, limited clustering | +0 |
| Thin | Under 5 pages, no clustering | -5 penalty |

---

## Overall Scoring (0-100)

| Component | Weight | Max Points |
|---|---|---|
| Experience | 25% | 25 |
| Expertise | 25% | 25 |
| Authoritativeness | 25% | 25 |
| Trustworthiness | 25% | 25 |
| **Subtotal** | | **100** |
| Topical authority modifier | | +10 to -5 |
| **Final score** | | **Capped at 100** |

Interpretation:

- **85-100** — Exceptional. Strong AI citation candidate across platforms.
- **70-84** — Good. Solid foundation; specific improvements will increase citability.
- **55-69** — Average. Multiple E-E-A-T gaps reducing AI visibility.
- **40-54** — Below average. Significant content quality and trust issues.
- **0-39** — Poor. Fundamental content strategy overhaul needed.

---

## Output Format

Write `GEO-CONTENT-ANALYSIS.md`:

```markdown
# GEO Content Quality & E-E-A-T Analysis — [Domain]
Date: [Date]

## Content Score: XX/100

## E-E-A-T Breakdown
| Dimension | Score | Key Finding |
|---|---|---|
| Experience | XX/25 | [One-line summary] |
| Expertise | XX/25 | [One-line summary] |
| Authoritativeness | XX/25 | [One-line summary] |
| Trustworthiness | XX/25 | [One-line summary] |

## Topical Authority Modifier: [+10 to -5]

## Pages Analyzed
| Page | Word Count | Readability | Heading Structure | Citability Rating |
|---|---|---|---|---|
| [URL] | [Count] | [Score] | [Pass/Warn/Fail] | [High/Medium/Low] |

## E-E-A-T Detailed Findings

### Experience
[Specific passages and pages with strong or weak experience signals]

### Expertise
[Author credentials found, technical depth assessment, specific gaps]

### Authoritativeness
[External validation found, topical authority assessment, gaps]

### Trustworthiness
[Trust signals present or missing, accuracy concerns if any]

## Content Quality Issues
[Specific passages flagged with reasons and rewrite suggestions]

## AI Content Concerns
[Low-quality AI content patterns detected, with specific examples]

## Freshness Assessment
| Page | Published | Last Updated | Status |
|---|---|---|---|
| [URL] | [Date] | [Date] | [Current/Stale/No Date] |

## Citability Assessment

### Most Citable Passages
[Top 5 passages AI platforms are most likely to cite, with reasons]

### Least Citable Pages
[Pages with lowest citability, with specific improvement recommendations]

## Improvement Recommendations

### Quick Wins
[Specific content changes that can be made immediately]

### Content Gaps
[Topics the site should cover to strengthen topical authority]

### Author / E-E-A-T Improvements
[Specific steps to strengthen E-E-A-T signals]
```
README.md

What This Does

Turns Claude into a content quality auditor that scores pages the way AI search engines do. It runs the E-E-A-T framework as a 100-point rubric with 28 named signals, then layers on word count floors, readability targets, paragraph structure rules, freshness scoring, and a topical authority modifier. Output is a written analysis with flagged passages and rewrite suggestions, not a number.

This one is about the writing itself: whether the content demonstrates real expertise and earns trust. Two neighbors cover the other halves of the same problem. geo-citability scores the structural properties that make a passage extractable and quotable. geo-platform-optimizer tunes the same site for each engine separately, since ChatGPT and Google AI Overviews pick sources differently.

From Zubair Trabzada's geo-seo-claude toolkit, packaged so a single CLAUDE.md gives you the full assessment.


Quick Start

Step 1: Create a Project Folder

mkdir -p ~/Documents/GEOContent

Step 2: Download the Template

Click Download above, then:

mv ~/Downloads/CLAUDE.md ~/Documents/GEOContent/

Step 3: Start Working

cd ~/Documents/GEOContent
claude

Give Claude a URL: "Score https://example.com/blog/our-guide on E-E-A-T." It fetches the page, scores the four dimensions, and writes GEO-CONTENT-ANALYSIS.md into the folder.


The Four Dimensions

Each is worth 25 points, scored against named signals rather than vibes.

Dimension What it measures Highest-weight signals
Experience First-hand involvement with the topic First-person accounts, original research, case studies with numbers
Expertise Knowledge depth and competence Visible author credentials, technical depth, stated methodology
Authoritativeness Recognition by others Inbound citations, press mentions, tier-1 publication history
Trustworthiness Publisher reliability Contact info, accurate claims, editorial or corrections policy, disclosures

The distinction between Experience and Expertise carries most of the weight. AI platforms separate content that reports on a topic from content written by someone who has done the thing. "We implemented this across 40 client sites and here is what broke" scores where "best practices for implementation" does not.


Word Count Floors

These are minimums for adequate coverage, not targets. Padding a 600-word answer to 2,000 words lowers the score.

Page Type Minimum Ideal Range
Homepage 500 500-1,500
Blog post 1,500 1,500-3,000
Pillar guide 2,000 2,500-5,000
Product page 300 500-1,500
Service page 500 800-2,000
About page 300 500-1,000
FAQ page 500 1,000-2,500

What Gets Flagged

The template ships a table of low-quality AI content signals that Claude checks each page against:

  • Generic phrasing ("In today's fast-paced world", "It's important to note that")
  • Perfect but empty structure: clean headings with shallow content beneath them
  • Hedging overload ("It depends on various factors") without naming the factors
  • Repetitive conclusions where each section restates the same point
  • Filler paragraphs that could be deleted with no information lost
  • Claims presented as fact with no attribution

AI-generated content is not itself a problem. Google clarified in March 2024 that production method does not matter as long as the result shows genuine E-E-A-T and human oversight. The flagged patterns are symptoms of thin content, whoever wrote it.


Freshness Scoring

Content without a visible date is treated as less trustworthy. Dates should be specific ("January 15, 2026"), not "recently", and should appear in both the rendered content and the structured data.

Last updated Status
Within 3 months Excellent
Within 6 months Good
Within 12 months Acceptable, may need refresh
12-24 months Warning, review for accuracy
No date, or 24+ months Critical, may be deprioritized

Content covering fundamentals that do not change gets marked evergreen and exempted, as long as it carries no time-dependent claims like "the latest" or "currently".


Topical Authority Modifier

After the 100-point E-E-A-T score, the site-wide coverage of its core topic adds or subtracts:

  • Authority (20+ pages, strong clustering): +10
  • Developing (10-20 pages, some clustering): +5
  • Emerging (5-10 pages, limited clustering): +0
  • Thin (under 5 pages, no clustering): -5

The assessment also looks for gaps: subtopics competitors cover that this site misses.


Tips

  • Score three or four pages, not one. The E-E-A-T signals that matter most are site-wide (author pages, contact info, editorial policy) and only show up when you look across a set.
  • Fix Trustworthiness first. Contact information, a privacy policy, valid HTTPS, and disclosures are 12 of the 25 points and take an afternoon.
  • The Experience gap is usually the real one. Adding one specific case study with numbers moves more than a week of rewriting.
  • Rerun after changes and diff the two analysis files. The score delta tells you whether the rewrite landed.

Limitations

  • Scoring is judgment-based, not deterministic. Two runs on the same page can land a few points apart. Use it for direction and priority order, not as a metric to report on.
  • Authoritativeness depends on external validation (press mentions, backlinks, Wikipedia presence) that Claude can only partially verify from a page fetch.
  • Readability targets are estimated from sentence and paragraph length rather than computed with a Flesch tool.
  • The word count benchmarks are English-language, general-web heuristics. Technical documentation and legal pages follow different norms.
  • The full geo-seo-claude suite runs 15 skills with Python scoring scripts and parallel subagents. This template is the content assessment on its own, driven by Claude reading the page.

$Related Playbooks