Content Quality & Review

When to Use

Use AI-powered quality review to ensure consistency, readability, and effectiveness across all content. Use the 4 C's framework (Clear, Concise, Compelling, Credible) for structured quality assessment.

Decision

Quality Criterion	What It Measures	Target Score	Action If Below Target
Clear	Readability, structure, jargon-free	70+	Simplify language, add structure
Concise	Word efficiency, focus	75+	Remove redundancy, tighten focus
Compelling	Engagement, relevance	70+	Improve opening, add examples
Credible	Accuracy, sources, expertise	80+	Add citations, verify facts

The 4 C's Quality Framework

Clear - Easy to understand by target audience - Free of jargon or unexplained technical terms - Logical structure and flow - Appropriate reading level

Concise - No unnecessary words or repetition - Focused on core message - Respects reader's time - Appropriate length for content type

Compelling - Engages reader interest - Addresses reader needs and pain points - Strong opening and closing - Actionable takeaways

Credible - Factually accurate - Properly cited sources - Author expertise evident - Professional tone and polish

Pattern

AI Content Advisor Implementation:

// Content analysis
$advisor = \Drupal::service('ai_content_advisor.analyzer');

$analysis = $advisor->analyzeContent($node, [
  'readability' => TRUE,
  'seo' => TRUE,
  'completeness' => TRUE,
  'engagement' => TRUE,
]);

// Results structure
$results = [
  'readability' => [
    'score' => 65, // Flesch Reading Ease
    'grade_level' => 8,
    'recommendations' => [
      'Reduce average sentence length from 24 to 15-20 words',
      'Replace complex words: "utilize" → "use"',
    ],
  ],
  'seo' => [
    'score' => 78,
    'keyword_density' => 1.2,
    'recommendations' => [
      'Add primary keyword to first paragraph',
      'Include 2-3 subheadings with keyword variations',
    ],
  ],
];

Automated Quality Scoring:

automator_type: llm_json_native_binary
field: field_quality_score
trigger: [create, update]

prompt: |
  Analyze this content using the 4 C's framework:

  Title: [node:title]
  Content: [node:body]

  Evaluate each criterion (0-100):

  1. Clear: Is it easy to understand?
  2. Concise: Is it appropriately brief?
  3. Compelling: Does it engage the reader?
  4. Credible: Is it trustworthy and accurate?

  For each, provide:
  - Score (0-100)
  - 2-3 specific improvement recommendations

  Output as JSON

Quality Gate Integration

Implement quality gates in publishing workflow:

/**
 * Implements hook_node_presave().
 */
function custom_quality_node_presave(NodeInterface $node) {
  // Run quality check before publishing
  if ($node->isPublished() && $node->isNew()) {
    $advisor = \Drupal::service('ai_content_advisor.analyzer');
    $quality = $advisor->assessQuality($node);

    // Minimum score to publish
    $min_score = 75;

    if ($quality['overall_score'] < $min_score) {
      $node->setUnpublished();
      \Drupal::messenger()->addWarning(
        sprintf(
          'Content quality score (%d) below publication threshold (%d). Saved as draft.',
          $quality['overall_score'],
          $min_score
        )
      );

      // Store quality report
      $node->set('field_quality_report', json_encode($quality));
    }
  }
}

Best Practices

Readability: - Target Flesch Reading Ease score of 60-70 for general audiences - Keep sentences under 20 words on average - Use active voice (target 80% or higher) - Break long paragraphs (max 4-5 sentences) - Add subheadings every 300-400 words

Fact-Checking AI Content: - Verify all statistics and data points - Check that cited sources actually support claims - Validate technical accuracy with subject matter experts - Test any code examples or instructions - Cross-reference with authoritative sources

Human Review Requirements: AI-generated content requires human review for: - Factual accuracy (AI hallucinates) - Brand voice and tone consistency - Legal and regulatory compliance - Sensitive topics (medical, legal, financial) - Competitive positioning and messaging

Common Mistakes

Wrong: Publishing AI content without human review → Right: AI hallucinates facts, misunderstands context, and produces plausible-sounding nonsense; always verify
Wrong: Using quality scores as strict gates without context → Right: A technically excellent how-to guide may score low on "compelling"; apply framework contextually
Wrong: Not updating quality criteria over time → Right: Readability targets, SEO requirements, and content standards evolve; review criteria quarterly
Wrong: Ignoring quality signals in analytics → Right: High bounce rate, low time-on-page, and poor engagement indicate quality issues; correlate with quality scores
Wrong: Over-optimizing for AI quality scores → Right: Writing to satisfy AI metrics can produce robotic content; optimize for humans first, metrics second