content

content

Covers the full content lifecycle: strategy, briefing, production (AI-assisted and human), and distribution. Applies across blog, email, UGC, and social formats.

1) Content brief template

# Brief: [Article Title]

**Target keyword**: [primary keyword]
**Search intent**: informational | transactional | navigational | commercial
**Word count**: [target range]
**Funnel stage**: TOFU | MOFU | BOFU

## Audience
- Who: [persona]
- Pain point: [specific problem this solves]
- Prior knowledge: beginner | intermediate | expert

## Structure
- H1: [exact title]
- H2s: [list of sections]
- Featured snippet target: [≤50 word answer to lead with]

## Must-include
- [ ] Cite: [source 1], [source 2], [source 3]
- [ ] Data point: [statistic or study]
- [ ] CTA: [specific action at end]

## Tone
[brand voice adjectives — e.g., "authoritative but approachable, no jargon"]

2) Topic clustering (programmatic)

from sklearn.metrics.pairwise import cosine_similarity
import numpy as np

def cluster_topics(keywords: list[str], embeddings: np.ndarray, threshold: float = 0.82) -> list[list[str]]:
    """Group semantically similar keywords into content clusters."""
    sim_matrix = cosine_similarity(embeddings)
    clusters, assigned = [], set()
    for i, kw in enumerate(keywords):
        if i in assigned:
            continue
        cluster = [kw]
        assigned.add(i)
        for j in range(i + 1, len(keywords)):
            if j not in assigned and sim_matrix[i][j] >= threshold:
                cluster.append(keywords[j])
                assigned.add(j)
        clusters.append(cluster)
    return sorted(clusters, key=len, reverse=True)

3) AI-assisted writing workflow

# Multi-pass writing: research → outline → draft → refine
async def write_article(brief: dict) -> str:
    # Pass 1: Research context
    research = await gather_research(brief["keyword"], brief["citations"])

    # Pass 2: Outline (structure-first)
    outline = await llm_call(
        system="You are a senior editor. Output only a structured markdown outline.",
        prompt=f"Create outline for: {brief['title']}\nResearch context:\n{research[:2000]}"
    )

    # Pass 3: Draft section by section (prevents context overflow)
    sections = []
    for section in parse_outline(outline):
        draft = await llm_call(
            system=brief["tone_instructions"],
            prompt=f"Write section: {section}\nBrief: {brief}\nPrior sections summary: {summarise(sections)}"
        )
        sections.append(draft)

    # Pass 4: Editorial polish
    full_draft = "\n\n".join(sections)
    return await llm_call(
        system="Fix grammar, improve flow, ensure E-E-A-T compliance. Do not change facts.",
        prompt=full_draft
    )