Content Brief Builder

Jobs-to-Be-Done is a framework from Clayton Christensen (Harvard Business School): customers don't buy products, they hire them to do a job. A functional job is a practical task ("compare several vendor options quickly"). An emotional job is about how they want to feel ("feel confident in the decision"). A social job is about how they want to be perceived ("present a well-researched recommendation to my team").

Most product content only addresses the functional dimension — which is exactly why the emotional and social dimensions are the larger content gap. AI retrieval works on semantic meaning, not keywords: content that addresses the feeling and the social context gets retrieved for queries that never state them explicitly, because the embedding captures the intent behind the phrasing.

The opening paragraph is the #1 citation zone — research on ChatGPT citation patterns has shown a substantial share of cited passages comes from the first 30% of a page. If the content type mismatches the reader's stage, the piece gets skipped regardless of how good the opening is. The opening itself follows the universal Answer-First requirements (Foundations §6.3); the journey stage determines which kind of content surrounds that opening.

Per Foundations §3.5, each journey stage calls for a distinct content type:

• Awareness — educational content; science foundations; problem explanation
• Consideration — comparison guides; expert reviews; testimonials; feature breakdowns
• Decision — authority signals; testimonials; scientific backing; thought leadership

Note on AIDA's "Interest" stage: the classic AIDA funnel splits the early funnel into Awareness and Interest. This tool uses the three-stage model from Foundations §3.5 because the two funnel stages produce the same content-type guidance; splitting them would not change the writing instruction.

Category Entry Points are the real-life situations that cause someone to think about a product category in the first place. The 7W framework (from Professor Jenni Romaniuk, Ehrenberg-Bass Institute) organises CEPs into seven dimensions: Who, What, When, Where, Why, With whom, How (emotional state).

A strong CEP-anchored piece opens by describing the situation before introducing the solution. This matches how real AI queries are phrased — situation-first, product-second. The piece becomes findable through situational queries, not just product queries.

Why multi-select: a single situation usually hits several CEP dimensions at once. "End-of-quarter review preparation" is simultaneously a WHEN (time-based trigger), a WHY (motivation: proving results), and a HOW (emotional state: pressure). Tagging all three gives the writer a richer frame and helps the piece rank against queries phrased through any of those lenses.

Skip this field entirely for foundational explainer pieces not anchored to a specific situation ("How does [category] work?").

In AI-driven retrieval, each H2 section of your page is effectively a standalone citation unit — AI systems frequently pull a single section rather than paraphrasing the whole page. A page with 6 self-contained H2 sections, each directly answering a real query, gives you 6 chances to be cited rather than 1.

Write queries exactly as a real person would type them: conversational, question-shaped, with the mix they actually use (problem queries, comparison queries, decision queries). These are your sentinel queries — the ones you'll track in ChatGPT, Perplexity, and Gemini to measure citation lift.

Aim for 3–6 queries for a full-length spoke. Hub pages usually have 1–2 queries, tightly branded. FAQ clusters can have 10+.

AI systems use entity recognition to decide what a page is "about." Content that refers to its subject generically ("our product", "this tool", "the [category]") gives AI no signal to connect the content to the underlying entity — so the content effectively becomes invisible to AI-based retrieval for queries about that entity.

The entity-first pattern is simple: on first mention, use the full entity name plus its category (e.g. "[Brand]'s [Product Name] [Category]"); on later mentions, use a shortened but still specific form (e.g. "the [Product Name]"); never use ungrounded generic references.

AI named-entity recognition identifies every entity on a page — products, technologies, components, people, competitors — and maps how they relate. A section about how the technology works should open by naming the technology; a section about a use case should open by naming the use case. An entity that's never named in full cannot be cited or connected.

Listing the supporting entities at brief stage does two jobs at once: it tells the writer which names to establish (and keep consistent — one canonical name per entity, every time), and it becomes the entity-architecture handoff the Technical stream needs to implement structured data. Entity decisions are made here, in the Content stream; schema implementation belongs to Technical.

The Princeton GEO study (Aggarwal et al., KDD 2024 — peer-reviewed) measured visibility lifts of up to +40% for the three strongest content methods: statistics, expert quotations, and citing credible sources. Fluency and technical-term optimization added +15–30%. The source-citation effect was strongest for lower-ranked pages — a #5-ranked page citing authoritative sources gained up to +115% visibility. (Evidence tier: Measured.)

These gains only materialise if the evidence is real — named experts, verifiable numbers, Tier-1 sources (peer-reviewed journals, government data, established industry research). And in the AI era a published claim doesn't fade: an AI answer restates it with the same confidence in month 18 as on day one, across every platform, stripped of your hedges. That cuts both ways — a strong sourced claim compounds, and an unsubstantiated one becomes a standing liability you can't recall.

The discipline is one rule: every number must trace to a primary record — the original document it came from (a test report, a platform dashboard export, an engineering spec sheet, a published study), not a blog that repeated it. If you can't name the record, the claim doesn't go in the brief.

Under FTC advertising-substantiation doctrine, every objective performance claim needs prior evidence — not evidence you plan to gather later. The four requirements:

• Pre-substantiation — the evidence must exist before the claim is published. "40% faster" needs a completed test showing 40%.
• Methodology match — the test must match the claim's conditions. Testing under one set of conditions doesn't cover a claim made about different conditions.
• Typical results — best-case figures require disclosing what typical users experience.
• Specificity — "up to X%" still requires data showing X% is achievable. "Up to" is not a loophole.

Knowing violations of FTC rules on deceptive practices carry civil penalties of up to $53,088 per violation (FTC Act §5, 2025 inflation adjustment, current as of 2026) — and "per violation" can mean per impression, per platform, per day. Claims in regulated categories (health, cosmetics, financial, environmental, country-of-origin) typically need legal or compliance review before publication, on top of substantiation.

An Answer Capsule is a 40–60 word, self-contained answer placed directly under the title or an H2 — with no links inside the capsule block itself. It names the entity, states the core claim, and includes one verifiable statistic. In a ~2M-session audit, 72.4% of ChatGPT-cited blog posts contained one — the single most common trait of cited content — and over 90% of cited capsules were link-free. Citations go in the supporting text below the capsule, not inside it.

The capsule is the executable format of the Answer-First principle: 44.2% of AI citations come from the first 30% of content (1.2M-answer analysis), and the start of a long context is the model's highest-attention zone ("Lost in the Middle"). Lead with the answer, in a block AI can lift whole. (Evidence tier: Correlational.)

From Bob Moesta's Demand-Side Sales 101 (2020), built on Christensen's JTBD research: every switch is governed by two forces pushing the reader forward (the push of the current frustration, the pull of the new solution) and two holding them back (the anxiety of switching, the habit of the status quo). Content that only sells the pull leaves three forces unaddressed — and the unaddressed anxieties are exactly what comparison-stage readers ask AI about.