The Brief Engine · Kelvico

I spent 8 hours on every semantic brief. Now I ship one in 15 minutes with 50x more depth.

This page is a full breakdown of the Kelvico Brief Engine. Built over 6 months. Trained across 7 verticals and 42 refinement sessions. Produced in partnership with the semantic SEO methodology pioneered by Koray Tuğberk GÜBÜR. Scroll for the manual workflow it replaces, what goes inside a single brief, and how to get a sample for your page.

Get a Sample Brief for Your Page Book a Strategy Call
Per brief
8 hours → 15 min
Monthly output
30 briefs → 300+
Per section
3 sentences → 13 fields
The Backstory

For two years I built semantic briefs by hand. One brief took me a full working day.

Before Kelvico, every semantic brief I produced was a day of manual assembly.

I would start with Ahrefs, pulling the full query dataset for the topic. Thousands of keywords. I would classify them into 5 query streams by hand in Google Sheets. Representative queries, sequential, correlative, boolean, implicit. One column per stream. Highlighting rows for hours.

Then competitor extraction. I would open the top 10 ranking pages in Chrome tabs. Read each one. Pull out entities into another sheet. PPR classification applied manually. Purpose attributes, property attributes, relationship attributes. A typical topic gave me 150 to 200 entities. Each one needed a row.

Then the Wikipedia ontology work. Semantic SEO done right requires mapping the entity taxonomy and the lexical relations around it. Synonyms, antonyms, hyponyms, hypernyms, co-hyponyms, meronyms. I would spend 45 minutes reading Wikipedia articles for a single topic, pulling relationship terms into my sheet.

Then Google Search Console. What queries are my client's pages already getting impressions for? Which ones convert? Which ones sit at positions 5 to 15 and could be pushed with the right brief? I would cross-reference GSC exports with the Ahrefs data. Still in Google Sheets. Still manual.

Then the SERP intelligence. For every target query, I would check the actual SERP. Featured snippets. People Also Ask boxes. Related searches at the bottom. Site links. The whole SERP feature inventory. Screenshots piled up.

Then I would open ChatGPT or Claude as a research assistant. Not to write anything. Just to help me map entity relationships that Wikipedia missed. I would paste my entity list and ask for unusual connections, contextual relationships, edge cases. Another 30 minutes per brief.

Then I would read everything the top 10 competitors had published on the topic. Not skim. Read. Pull out which frames they filled and which they ignored. Which frames none of them filled (the gap opportunity). Mark it all in the sheet.

Then the brief structure. I would take all of that research and try to turn it into a heading outline. Queries mapped to headings. Frame coverage planned. Section order decided. Dedup tracking initialized. Bold system rules set. Format diversity plotted.

And then the last hour. Brand positioning layer. Conversion context. Voice calibration notes. The human glue that makes the brief usable.

Total time per brief, start to finish, by a trained operator who had been doing this for two years. Seven and a half to eight hours.

My maximum output was 30 briefs per month. Any team member I trained up capped out around 15 briefs per month before quality dropped. Training took 6 to 8 weeks per hire. Every new operator produced slightly different quality. Every brief depended on the individual.

And the briefs still had problems. Source context disconnected from queries. Brand positioning bolted on as an afterthought. Conversion angle missing. Pages that ranked technically but did not convert. It felt like a research document, not a production spec.

That was the ceiling. 30 briefs per month. $2,400 per brief of my time. And the ceiling was permanent. I could not scale it by hiring. I could not shortcut it by prompting. The work was the work.

So I spent 6 months building the engine that does it for me.

The Positioning

This is not a tool. It is a 12-engine production system with the brief engine at the center.

The Brief Engine is one of 12 engines in a Kelvico content system. It is the most important one because it is where all the intelligence comes together into a production spec a writer, human or AI, can execute against.

It is not a single prompt. It is not a template you fill in. It is not a Claude project with one instruction document. It is a multi-phase pipeline that runs four engines before it even gets to the brief.

Phase 1

Research Engine

Extracts 6 to 10 layers of competitor intelligence automatically.

Phase 2

Outline Engine

Maps every target query to a heading. Fills all 9 semantic frame slots. Produces the architectural skeleton.

The Engine
Phase 3

Brief Engine

Layers the 13 production fields per section onto the outline. Entity maps. Dedup tracking. CTA guidance. Modality.

Phase 4

Verification Engine

Fact-checks every claim before the brief is finalized.

The brief that comes out the other end is not a 3-sentence outline. It is a 13-field production spec per section, built on semantic architecture principles derived from Koray Tuğberk GÜBÜR's framework and operationalized through 42 refinement sessions across 7 client verticals.

One brief now takes 15 minutes of compute time plus about 10 minutes of human review. Same operator. Same quality standard. 50 times the output.

Before the Engine

Every step a semantic brief requires, done by hand.

Below is the exact workflow I used to produce a single brief manually. If you are doing semantic SEO correctly, you recognize this list. If any of these steps are missing from your current process, your briefs are not actually semantic briefs.

01

Query dataset extraction

Pull the full query set from Ahrefs. Export to sheet. Thousands of rows per topic.

30 min
02

5-stream query classification

Classify every query by intent type. Representative, sequential, correlative, boolean, implicit.

60 min
03

Competitor content extraction

Read the top 10 ranking pages. Extract entities, headings, trust signals, voice patterns into the sheet.

90 min
04

Entity map with PPR classification

Map every entity found. Classify by Purpose, Property, Relationship. 150 to 200 entities per topic.

60 min
05

Wikipedia ontology research

Pull synonyms, antonyms, hyponyms, hypernyms, co-hyponyms, meronyms. Read the source articles.

45 min
06

Search Console cross-reference

Match the query set against existing impressions and rankings. Identify positions 5 to 15 to push.

30 min
07

SERP feature audit

Capture featured snippets, PAA, related searches, site links for every target query.

30 min
08

Heading-to-query mapping

Map every classified query to a specific heading or subsection. No orphan queries.

45 min
09

Frame coverage planning

Ensure all 9 semantic frames are filled. Identify which frames no competitor has filled.

45 min
10

Format diversity plotting

Decide which sections use tables, lists, callouts, card grids, prose.

30 min
11

Brand positioning layer

Add voice calibration, conversion angle, CTA integration per section.

30 min
12

Human assembly and review

Compile everything into a production brief. Review for coherence. Last-mile polish.

90 min
Total per brief
7 hours 45 minutes
Trained operator. Two years of practice. Still capped at 30 briefs per month before quality dropped.
If you are producing briefs faster than this without automation, you are skipping steps. And if you are skipping steps, your briefs are not semantic briefs. They are outlines with keywords.
After the Engine

Same 12 steps. Running in 15 minutes through the Kelvico pipeline.

The engine does not shortcut any of the 12 steps. It runs all of them. The difference is they run in sequence through automated engines, not through a human in Google Sheets.

Research Engine absorbs

Steps 1 through 7

01 02 03 04 05 06 07

Query extraction, classification, competitor content intelligence, entity mapping with PPR, Wikipedia ontology research, Search Console cross-reference, and SERP feature audit all happen as parallel sub-processes.

Outline Engine absorbs

Steps 8 and 9

08 09

Heading-to-query mapping and frame coverage planning. Every heading tied to its classified queries. Every frame slot verified filled.

Brief Engine absorbs

Steps 10, 11, 12

10 11 12

And expands them into 13 production fields per section. Format, brand positioning, and assembly become structured decisions, not narrative hand-holding.

The 13 fields the Brief Engine computes per section

Each field is a decision the engine makes using the upstream research

01 question_modality Which of 20 question-answer patterns this section follows
02 gap_context What the top competitors missed on this topic
03 frame_context Which of 9 semantic frames this section fills
04 role_guidance Voice, persona, tone markers for the section
05 format_guidance Table, list, callout, card grid, or prose
06 dedup_tracking Which proof points to deploy here
07 semantic_requirements Entities, lexical relations, variants to include
08 bold_guidance Which knowledge graph triples to bold
09 content_boundaries What the section must NOT include
10 persona_focus Which customer persona cares most about this section
11 cta_guidance CTA type, placement, phrasing
12 trust_signals Which certifications, numbers, named entities to deploy
13 objection_handled Which buyer objection this section resolves

The writer receiving the brief does not guess. Every sentence has a job. Every decision has already been made upstream.

Typical AI brief

3 sentences. "Write a 1,500-word article on X. Include keywords Y and Z. Use headings H1 through H6."

Kelvico brief

13 fields per section, multiplied by 8 to 15 sections. 104 to 195 structured decisions per brief. Compiled automatically. Delivered in 15 minutes.

Proof Through Contrast

Same section. Same topic. Two very different briefs.

Below are two briefs for the same section of the same article. Topic. "How does email deliverability actually work for B2B SaaS."

Typical AI-generated brief

Write a 400-word section explaining how email deliverability works. Include keywords. email deliverability, SPF, DKIM, DMARC. Use an H2 heading. Add a bullet list if possible. Make it engaging.

Zero entity mapping Zero query classification Zero frame context Zero modality specification Zero dedup tracking Zero content boundaries The writer is guessing
Kelvico Brief Engine output
SECTION_ID: s2-mechanics
HEADING: How does email deliverability actually work for B2B SaaS?

question_modality: MECHANISTIC
  // expected response opener:
  "Deliverability works through three mechanical inputs..."

gap_context: Top 5 competitors all name SPF/DKIM/DMARC but
  none quantify how each input contributes to inbox
  placement. This is our opening.

frame_context: MECHANISM frame (one of 9 required).
  Reader state. already knows deliverability is a
  thing, wants to understand the mechanics.

role_guidance: Technical but accessible. Senior
  engineer talking to senior marketer. No fluff.

format_guidance: 3-paragraph structure. First opens
  with a declaration. Second breaks into 40/45/15
  percentage weights. Third recommends where to
  start.

dedup_tracking: FULL deployment of the 40/45/15
  authentication/reputation/content breakdown. Do
  not repeat in sections s4, s6, or s8.

semantic_requirements: Must name these 6 entities.
  SPF, DKIM, DMARC, complaint rate, bounce rate,
  spam trigger words. Lexical relation. "inbox
  placement" as synonym for "deliverability" at
  least once.

bold_guidance: Bold the knowledge graph triple in
  the declaration. Format. "Deliverability is a
  function of three mechanical inputs."

content_boundaries: Do NOT discuss list building,
  segmentation, or copy optimization. Those are
  covered in sections s5 and s7.

persona_focus: Marketing Operations Lead at a
  50-500 employee SaaS. Technical enough to
  understand DNS records.

cta_guidance: Soft CTA at paragraph 3 end. Anchor.
  "see our authentication audit checklist" linking
  to s8.

trust_signals: Reference Gmail's February 2024 bulk
  sender requirements as the timely anchor. Cite
  the 0.3% complaint rate threshold.

objection_handled: "Does fixing authentication
  actually move the needle?" Answer in declaration.
6 named entities required Frame specified (Mechanism) Modality specified Dedup tracking in place Content boundaries drawn Persona locked CTA integrated Trust signal with date Objection named
This is what a semantic brief is supposed to look like. Most teams cannot produce this manually at volume. The engine does it in 15 minutes.
The Differentiation

If you have used a generic AI brief tool, you know the ceiling. Here is how Kelvico goes past it.

Every AI brief tool on the market today works the same way. One prompt, one model call, one output. The output quality is capped by whatever the prompt engineer could cram into the system message.

Kelvico is fundamentally different because the Brief Engine is the 8th engine in a 12-engine pipeline. Every engine before it has already done specialized work. The Brief Engine does not have to research, classify queries, extract entities, or map competitor gaps. All of that is already done and handed to it as structured input.

Capability
Generic AI Brief Tool
Kelvico Brief Engine
Source of intelligence
One LLM pass on the topic
7 upstream engines producing structured data
Entity map depth
None or surface-level
150-200 entities with PPR classification
Query classification
Keyword list, no intent structure
5-stream classification with modality mapping
Frame coverage
Usually 2-3 of 9 frames filled
All 9 frames filled and verified
Fields per section
2-5 generic instructions
13 structured fields
Dedup tracking
None
Explicit proof point allocation across sections
Brand voice calibration
Generic tone setting
6 tone profiles with DO and DO NOT vocabulary
Verification
None
Mandatory fact-check engine before brief ships
Custom per client
One-size-fits-all
Built from your competitors, queries, brand
Methodology
Proprietary prompt
Koray Tuğberk GÜBÜR framework, productized
A generic AI tool produces a brief in 90 seconds. It feels fast. What you are actually getting is a 2-field brief dressed up as a 13-field brief. Writers following it still guess. Results still vary. Kelvico is slower on paper and faster in practice because the writers do not have to guess.
Under the Hood

What gets computed for one section of one brief.

Expand each field below to see what the engine produces for a real section. The sample section we are showing. "Cost and pricing" in a SaaS comparison page.

Sample section. Cost and pricing · SaaS comparison page
01 question_modality
Engine detects the section heading asks a "how much" question. Assigns modality. QUANTITATIVE. Required response opener pattern. "[Product X] costs [price] per [unit] at [tier level]."
02 gap_context
Engine reads competitor pricing sections. Detects that 4 of 5 competitors list tier prices but none explain pricing mechanics (per-user vs per-seat vs usage-based). Gap identified. Section job. explain the mechanic first, then list the numbers.
03 frame_context
COST frame (one of 9). Reader state. already knows the product exists, already knows the features, now evaluating affordability.
04 role_guidance
Transparent pricing operator voice. No hedging. No "contact us for pricing" language. Zero marketing fluff. Treat the reader as a buyer comparing specifically.
05 format_guidance
Pricing table with 4 columns. Tier name, monthly price, key feature threshold, best-for scenario. Below table, 2 paragraphs explaining mechanics and hidden costs.
06 dedup_tracking
Full deployment of per-seat pricing breakdown here. Short-form references only in sections s4 (comparison table) and s9 (FAQ). Do NOT restate the pricing mechanic in section s6 (use cases).
07 semantic_requirements
Must name. monthly fee, per-seat cost, annual discount, setup fee, overage charges, usage tier breakpoints, trial terms. Lexical coverage. "subscription" and "plan" as synonyms. Antonym. "usage-based" as contrast to "flat-fee."
08 bold_guidance
Bold the declaration sentence's knowledge graph triple. Format. "[Product] charges [X] dollars per seat per month on the [tier name] plan." Also bold the "best-for scenario" value in each pricing table row.
09 content_boundaries
Do NOT discuss feature comparisons in this section (section s4). Do NOT discuss ROI math (section s7). Do NOT deploy customer testimonials (section s5). This section is strictly about the mechanics and numbers.
10 persona_focus
Primary. finance-approving buyer at a 100-500 employee SaaS. Secondary. mid-market procurement lead. NOT the end user of the software. Pricing language should match procurement-approved terms.
11 cta_guidance
Mid-section soft CTA after the pricing table. Anchor text. "see how the ROI breaks down" linking to section s7. Primary CTA is deferred to final section.
12 trust_signals
Reference. specific pricing page URL from the product's site. Reference. G2 average rating across 200+ reviews. Reference. NAID AAA or equivalent certification if applicable. Deploy 2 of these. Save 3 others for later sections.
13 objection_handled
"Is this pricing actually transparent or am I going to discover hidden costs?" Answer this in the mechanics paragraph by naming every fee category the buyer could encounter.
This is ONE section. A typical brief has 8 to 15 sections. The Brief Engine produces 13 fields for each of them in 15 minutes. That is between 104 and 195 structured decisions per brief, computed automatically, handed to the writer on a plate.
Track Record

What 6 months of engine building produced.

42
Refinement sessions building and improving the engine
1,000+
Briefs produced using this engine in production
0
Briefs that required post-shipping restructure for architectural completeness
7
Client verticals tested. SaaS, ecommerce, healthcare, finance, real estate, B2B services, affiliate
13
Structured fields computed per section
9
Semantic frames verified on every page
5
Query streams classified automatically
6
Lexical relation types mapped per topic
3,000+
AI signature patterns checked before shipping
k.
Every refinement made the next brief sharper. Every vertical exposed edge cases the engine now handles. The system compounds. A brief produced today carries 42 rounds of accumulated intelligence that no single operator could maintain in memory.
Use Cases

Five page types where this brief engine has shipped production briefs.

01

SaaS Informational Guides

Long-form semantic content for mid-market SaaS companies. Brief engine produces briefs for 5,000-word pillar pages with 10-15 sections and full frame coverage.

02

Ecommerce Product Pages

Product page briefs with fixed and variable heading architecture. Includes contextual comparison tables, scenario-based recommendations, and named staff CTAs.

03

Affiliate Review Pages

Multi-entity review briefs with mandatory verification fields. Each entity, product, company, platform, gets its own structured brief section.

04

Healthcare Service Pages

Regulated content briefs with compliance woven into section-level guidance. Patient-facing language tested for clarity.

05

B2B Comparison Pages

Honest comparison briefs with conditional verdicts. Worth it IF, not worth it IF. Competitor gap analysis built into the brief guidance.

Try It

See what the Brief Engine produces for your page.

Fill in the brief request form. You will receive a full semantic brief within 48 hours, produced by the Kelvico Brief Engine. No charge. No commitment. If you would rather talk first, book a 30-minute call instead.

We personally review every request. Response within 48 hours.
Option B opens our Calendly. Option A emails us at hamid@kelvico.ai with your details.
We personally review every brief request. If your page is not a fit for the engine, wrong industry, pre-revenue, etc., we will tell you honestly rather than send a generic sample.
Questions

Eight questions we hear most often from semantic SEO practitioners.

Is this just a wrapper around ChatGPT?
No. The Brief Engine is one of 12 engines in a sequential pipeline. Before the Brief Engine runs, seven other engines have already executed specialized work. competitor extraction across 6-10 layers, entity mapping with PPR classification, query classification across 5 streams, Wikipedia ontology research, SERP feature audit, heading-to-query mapping, and frame coverage planning. The LLM that handles the final briefing pass receives structured input from all of those, not a blank prompt.
Whose methodology is this built on?
Koray Tuğberk GÜBÜR's semantic SEO framework is the foundation. PPR entity classification, 5-stream query networks, frame semantics, lexical coverage, contextual borders. All of it maps to his work. Kelvico is the productized operationalization of that methodology, not a replacement for it. If you are new to semantic SEO, start with Koray's training. If you are already trained and tired of doing it by hand, Kelvico is the engine.
How is the entity map built?
The Research Engine extracts entities from 6-10 competitor pages per topic, classifies each entity by PPR (Purpose, Property, Relationship), cross-references with Wikipedia ontology for lexical relations, and verifies entity existence through independent source checking. The final entity map for one topic typically contains 150-200 entities with classifications and relationships.
Can I run this myself on my own content?
Yes. Kelvico builds the engine as a custom system for your business. You own it. We deliver the engine plus training for your team. Most clients choose to have us run the engine as a managed service, but self-operation is the default contract.
How is this different from Frase, Surfer, Clearscope, or other content brief tools?
Generic content brief tools do keyword-density analysis and simple heading suggestions. They do not do PPR entity classification, 5-stream query networks, 9-frame coverage verification, modality matching, dedup tracking, or 13-field structured briefs. The output of a Kelvico brief is a production specification, not a content optimization suggestion.
Do you train the engine on our brand voice?
Yes. The Tone and Voice Profile Engine runs after the Brief Engine and calibrates 6 tone profiles per client based on competitive voice gap analysis. Your engine knows how your brand speaks, what words to avoid, what phrases to emphasize. This is baked into every brief before it ships.
What is the average turnaround time?
Engine build from first call to first production brief. 2 to 3 weeks. After that, a single brief takes 15 minutes of engine compute plus 10 minutes of human review. A client team running the engine can produce 20 to 40 briefs per week without quality degradation.
What industries will you not build this for?
Gambling, predatory lending, health misinformation, adult content, and anti-science categories. Every other legitimate vertical is welcome. We have built engines for 7 verticals so far.
Your Move

Get a brief for your page.

One form. Two options. We respond within 48 hours.

Get Your Brief
Or reply to any LinkedIn post that brought you here. We read every one.
Jump to brief form