CONTENT ENGINE

Six stages. From portfolio audit to published page.

Not a blank-canvas writer. A six-stage content operations system. Classify your full portfolio, recommend optimize/create/retire with locked precedence, generate 9-section briefs, ship drafts only after they pass the 8-dimension quality gate at 60, cluster the taxonomy, and calibrate weights against day 7/14/30 outcomes. Growth and Agency only.

§ 01Why it exists

A blank-canvas writer is the wrong tool. A pipeline is the right one.

Every other AI content tool starts with a blank canvas. You type in a topic. It writes 800 words. You publish. Six months later you have 50 pages competing for the same query, three duplicates of your existing landing page, and no idea which ones actually drive AI mentions.

The Content Engine starts somewhere else. Stage 1 classifies every page you already have as winning, opportunity, underperforming, dormant, or neutral. Stage 2 recommends optimize, create, or retire with a locked precedence so you can't end up with conflicting suggestions on the same topic. By the time the engine generates a single draft, it knows what you already have and what's missing.

Stage 4 is where most other tools fail. The Content Engine refuses to deliver a draft that scores below 60 on the eight-dimension quality gate. Citation likelihood, intent match, schema validity, keyword coverage, internal links, length, readability, originality. If the composite is below 60, the API returns HTTP 422 and you get a focus-dimension hint to regenerate against — never a bad draft to clean up.

Stage 5 clusters your full content portfolio with DBSCAN into 11 content types and computes the internal link map. Stage 6 measures each published piece at day 7, day 14, and day 30, and calibrates the brief weights monthly with a ±3pp cap so the engine learns your brand without overfitting. It is not a writing tool. It is a content operations system.

§ 02The six stages

Sequential. Each gates the next.

CE-1
Portfolio classification

Every page on your site is classified daily into one of five verdicts: winning, opportunity, underperforming, dormant, or neutral. The classifier reads tracking signals, GSC data, GA4 sessions, and AI mention frequency. You start every recommendation cycle knowing what you have.

CE-2
Optimize / create / retire

Each portfolio page gets a recommendation: optimize, create, or retire. Dedup precedence is locked retire > optimize > create — you'll never get conflicting suggestions for the same topic. Recommendations refresh weekly on Monday 8 AM IST.

CE-3
9-section briefs

Each create or optimize recommendation generates a structured 9-section brief: target query, intent, audience, structure, key facts, schema, internal links, anchor variants, and tone. Briefs go through draft → approved before content generation. You review the brief, not the draft.

CE-4
Quality-gated generation

The LLM generates HTML + JSON-LD from an approved brief. The output is scored on 8 locked dimensions; composite must be ≥60 to pass. Drafts that fail return HTTP 422 with a focus-dimension hint for regeneration. You never receive bad content — you receive content that passed the gate or a clear explanation of why it can't be generated.

CE-5
Taxonomy + internal linking

DBSCAN-clusters your full content portfolio into 11 content types. Computes an internal link map across the cluster and recommends specific anchor text for every interlink. Internal links improve both human navigation and AI extraction; this stage makes sure new content joins the network correctly.

CE-6
Performance loop + calibration

Baseline + day 7 + day 14 + day 30 snapshots on every published piece. AI mentions (0.35), GSC clicks (0.25), GSC impressions (0.15), GA4 sessions (0.15), GA4 conversions (0.10). Monthly weight calibration capped at ±3pp with 20-sample minimum so the engine learns your brand's actual conversion patterns without overfitting.

§ 03Capabilities

What separates this from a prompt template.

Knows your full portfolio

CE-1 classifies every page on your site. The engine doesn't generate in a vacuum — it generates with full knowledge of what you already have, what's winning, and what's dormant.

Hard quality gate at 60

Eight dimensions, locked weights, hard cutoff. Drafts that score below 60 return HTTP 422 — you don't get bad content. The gate is the floor, not a guideline.

Locked dedup precedence

retire > optimize > create. The engine refuses to recommend a new page when an existing page should be optimized first, and refuses to recommend optimize when a page should be retired. No conflicting suggestions on the same topic.

Embedded JSON-LD on every page

FAQPage, Article, Product, BreadcrumbList — generated and validated before delivery. Schema validity is one of the eight quality dimensions; a draft can't pass the gate without correct schema.

DBSCAN-clustered taxonomy

Stage 5 clusters your content into 11 categories and computes the internal link map. Every new piece you publish joins the cluster correctly with anchor-text suggestions tuned to the existing network.

Self-calibrating weights

CE-6 measures actual outcomes at day 7, 14, and 30 and adjusts the brief-stage weights monthly. ±3pp cap and 20-sample minimum prevent noise from moving the dials. The engine learns your brand without overfitting.

§ 04Who it is for

Teams without a three-person writing staff.

Marketers without a content team

You know which pages you need but don't have writers to produce them. The engine handles the draft and the structure; you handle the brand voice review on briefs that already passed quality.

SaaS companies targeting category queries

When buyers ask AI "what is the best [category] tool," you need comparison and definition pages structured for AI extraction. CE-1 finds the gaps in your portfolio and CE-2 prioritizes which to fill first.

Agencies scaling content production

Run the full six-stage pipeline across multiple client brands without multiplying your writing team. The quality gate guarantees no client receives a draft that wouldn't pass internal review.

Teams measuring content ROI

CE-6 baselines every published piece and snapshots performance at day 7, 14, and 30. You can track how AI mentions, GSC traffic, and conversions moved after each piece — measured against your brand's own history, with a volume floor so thin samples are flagged as inconclusive rather than overstated.

§ 05Pricing

Included on Growth and Agency.

Growth
$149/mo

Full six-stage pipeline. Portfolio classification, recommendations, briefs, generation, taxonomy, performance tracking.

Agency
$249/mo

Everything in Growth + Quality Calibration UI access (CE-6 weight tuning per client brand).

See full pricing →
§ 06Common questions

Content Engine questions.

What's the quality gate? What happens if a draft fails?

Every generated draft is scored across eight locked dimensions: citation likelihood (0.25), intent match (0.20), schema validity (0.15), keyword coverage (0.15), internal links (0.10), length appropriateness (0.05), readability (0.05), and originality (0.05). The composite score must be ≥60 to pass. If a draft scores below 60, the API returns HTTP 422 and the engine refuses to deliver it. You don't get bad content — you get a regeneration with a focus dimension hint, or a clear explanation of why generation isn't possible for this brief.

How is this different from ChatGPT or Jasper?

ChatGPT and Jasper are blank-canvas writers. The Content Engine is a six-stage pipeline that knows your portfolio. Stage 1 classifies every page on your site as winning, opportunity, underperforming, dormant, or neutral. Stage 2 recommends optimize/create/retire actions with a locked precedence (retire > optimize > create) so you don't get conflicting suggestions. Stage 3 produces a 9-section brief that goes through approval before generation. Stages 5 and 6 cluster your content with DBSCAN and track performance at day 7, 14, and 30. It's not a writing tool — it's a content operations system.

What if I already have content for a topic?

Stage 2's dedup precedence handles this exactly. If a portfolio page exists for the topic, the engine recommends optimize, not create. If the page is a dormant duplicate, it recommends retire. If the topic genuinely has no coverage, it recommends create. You won't end up with two pages competing for the same query because the engine surfaces the conflict before the brief is generated.

How does the day 7/14/30 performance loop work?

Stage 6 captures baseline metrics when content publishes (AI mentions, GSC clicks, GSC impressions, GA4 sessions, GA4 conversions). At day 7, 14, and 30, it snapshots the deltas. Monthly, it calibrates the brief-stage weights against the actual outcomes — but with a hard ±3 percentage point cap and a 20-sample minimum so noise can't move the dials. The result: the engine gets better at predicting what will work for your specific brand over time, without overfitting to a few outliers.

Are there legal concerns with AI-generated competitor comparison pages?

The engine generates balanced, factual comparison content. Every comparison page includes pros and cons for both brands, uses publicly available information, and passes the schema_validity dimension of the quality gate before delivery. The format follows the same structure that editorial review sites use. You own the content and can edit it before publishing.

Who owns the content?

You own everything. There are no licensing restrictions, attribution requirements, or usage limitations. Once generated, the content is yours to publish, edit, redistribute, or repurpose however you want.

See which pages are missing.

Run a free Scorecard to find the AI queries where you have no content. The Content Engine turns those gaps into briefs, drafts, and quality-gated published pages.

Check your visibility, free →
60 seconds · no credit card · no signup for first result