The Problem with Generic AI Content
Default AI content:
- Fails NLP 60/40 balance (reads like a keyword list)
- Contains AI stop words ("delve", "crucial", "navigate")
- Lacks entity specificity (vague descriptions, no numeric data)
- Misses contextual bridges between sections
- Has no author voice or experience markers
These frameworks solve each problem systematically.
Framework 1: SEED + Growth (Author Persona)
Before writing any SEO content with AI, establish an author persona that the AI will write as.
SEED prompt — entity research:
For the niche "[niche]", provide:
1. 60 sub-topics, comma-separated
2. 5 famous writers or publications in this niche
3. 10 professional certifications related to this niche
4. 5 historical people, places, or events connected to this niche
5. 10 specific statistics about this niche
6. 5 regulatory associations or governing bodies
Growth prompt — persona construction:
Using the research above, create a professional persona for an author in [niche] with these specifications:
- Background: [industry-relevant origin story with E-E-A-T signals]
- Education: [relevant degrees + expertise in all 60 sub-topics from SEED]
- Career: [specific positions at named [niche]-relevant businesses]
- Publications: [named articles, guides, industry publications]
- Personal: [humanizing details connecting to expertise]
- Currently: [active role at a [niche] business in [city], writing to help people make informed decisions]
Rules: Never mention writing background. Position as practitioner, not writer.
Framework 2: Intent-First Prompting
Always declare the search intent and audience before asking for content:
Write for someone who searched "[exact keyword]" on Google.
Their hidden intent is [hidden intent from intent-analyzer].
They are at [early research / comparison / ready to buy] stage.
They need [specific information gap they're trying to fill].
Do NOT write for someone who already decided to [action]. Write for someone who is evaluating whether to [action].
Framework 3: Entity-Seeded Content Brief Prompt
Injects entity requirements directly into the prompt:
Write a [content type] about "[topic]" for a [business type] in [city, state].
Required entities (must appear in the content):
- Primary: [list 5-8 primary entities]
- Supporting: [list 5-10 supporting entities]
- Locations: [city, neighborhoods, landmarks]
- Trust: [license type, certification name, association name]
Required data points:
- Include cost range: $[X] to $[Y]
- Include timeline: [X] to [Y] [units]
- Include at least 2 statistics from authoritative sources
Required formats:
- H2: [question 1], [question 2], [question 3]
- Featured snippet target on H3 "[specific question]" (40-60 words, declarative, entity-first)
- FAQPage: 5 questions minimum
Framework 4: Anti-AI-Tells Refinement Prompt
Apply after any AI-generated draft:
Review this content and eliminate all AI-generated tells:
1. Find and replace these banned phrases: delve, crucial, navigate, multifaceted, embark, "in today's digital landscape", "it's worth noting", "having said that", "not only that", "a plethora of"
2. Find any sentence starting with "It is" or "There are" — rewrite as active voice
3. Find all vague quantifiers ("several", "many", "a few") — replace with specific numbers
4. Find any paragraph that could be removed without losing specific information — remove it
5. Add one first-person experience marker per section: "When we [specific action] on a [specific project type] in [location]..."
Return the revised content only.
Framework 5: Featured Snippet Chain
Three-prompt chain to engineer featured snippets:
Prompt 1: Generate question-format H3 for this topic that matches how users search Google.
Prompt 2: Write a 40-60 word declarative answer to "[H3 question]" that: starts with the entity name, contains no qualifiers (typically, usually, may), includes one numeric data point.
Prompt 3: Expand the answer into a full section (200-300 words) that supports the 40-60 word snippet without contradicting it.
Framework 6: Schema-Ready Content Prompt
Produces content already formatted for FAQPage extraction:
Write a FAQ section for "[page topic]" with 6 questions and answers.
Format requirements:
- Each question: H3 format, natural language, how a person would type it in Google
- Each answer: 40-200 words, starts with entity name, declarative, no promotional language
- Ensure each question-answer pair is standalone (readable without context)
Output as: Q1, A1, Q2, A2... format for easy copy-paste into FAQPage JSON-LD
Cross-References
koray-content-brief— semantic brief that these prompt frameworks execute againstcontent-mastery— quality standards that AI-generated content must meeteeat-audit— E-E-A-T verification of AI-generated content before publishingai-llm-seo— author persona framework for AI content that maximizes AI citation
#seo-sop #seo #prompting #ai-content #content-production #frameworks