Getting Started

How It Works

One repeating loop, made of three workflows. Each turn of the loop makes the store a little stronger — and a person signs off before anything goes live.

1. Understand. The system reads your catalog, every US page, what people search for, where you rank, and any first-hand data you've supplied.
2. Decide. It surfaces the highest-value opportunities — filtered to the complete systems you actually sell.
3. Produce. It writes a new page, or improves an existing one, grounding every spec and safety claim in your data and in code.
4. Check. The piece passes the guardrails — accurate claims, safety quoted verbatim, no duplicate-page conflicts, AI-search quality.
5. You publish. You receive a ready-to-paste package; your team puts it live in Magento.
6. Measure. The system tracks what moved and feeds that back into the next decision.

What keeps it accurate

Three foundations sit under everything:

• Product Intelligence — a structured copy of your catalog (diameters, wall types, fuel compatibility, certifications) so the system never invents a spec.
• Canonical safety paragraphs — the exact text of the US standards (NFPA 211, UL listings, IRC) is stored once and quoted word-for-word, so fire-safety content is never paraphrased or guessed.
• Your unique inputs — any first-hand information you supply (real install data, test results, proprietary specs, field experience, photos) is folded directly into the content. See Content Engine.

Getting Started

What You Get

Concrete deliverables on a weekly rhythm — plus a system that keeps improving in the background.

The publish-ready package

Every content piece arrives as a single document your team can implement directly:

• Target & secondary keywords, search intent, and SERP-feature targets
• Title tag, meta description, URL, and headings
• Full body copy, ready to paste
• Internal links in and out, with anchor text
• Validated structured-data (schema) block
• Image specs and alt text
• Publishing checklist (canonical, indexing, sitemap, redirects) and a duplicate-page clearance

Modules

The 6 Systems

The engine is built from six parts that work together. Status reflects the current trial stage and updates weekly.

Modules

Content Engine

Two of the system's three workflows live here: Topic Generation (what to write) feeds Content Generation (writing it). The third workflow — Content Refresh — improves pages you already have.

Workflow 1 — Topic Generation

In: your rankings, catalog, competitors, and search demand.  Out: a prioritised list of topics, each routed to a new page or a refresh.

1. Gather candidates from several angles at once — gaps in what you rank for, demand your catalog can serve, competitor coverage, the full tree of related questions, your on-site search log, and queries where an AI Overview appears but you aren't yet cited.
2. Apply the complete-systems filter — drop anything you don't sell as a full system (standalone caps, flashings, single accessories).
3. Score what remains by value and winnability.
4. Route each topic to a new page (Workflow 2) or an improvement to an existing page (Workflow 3).

Workflow 2 — Content Generation

In: a chosen topic.  Out: a finished, quality-checked, publish-ready page.

1. Analyse the search results for the target query — what wins, and what format Google rewards.
2. Build a brief — keywords, the questions to answer, which products to link, the exact safety code to cite, and the outline.
3. Incorporate your inputs — any unique data you've supplied is woven in (see the callout below).
4. Draft — answer-first, easy to scan, every fact grounded in your data.
5. Enrich — add internal links and validated schema.
6. Gate — pass the claim, safety, quality, and duplicate-page checks.
7. Package & hand over — you publish; we measure.

Modules

Content Refresh

The third workflow. The system watches your existing pages and improves the ones that need it — before they quietly lose ground.

In: signals that a page is slipping.  Out: a targeted, quality-checked refresh with before/after tracking.

The system notices when a page slips

Eight signals continuously flag pages worth revisiting — including:

• A tracked keyword drops in the rankings
• A competitor enters the top results for one of your terms
• An AI Overview appears on a query where you rank but aren't cited
• Click-through or freshness softens over time
• Two of your pages start competing for the same query
• A page is found carrying a claim that fails our accuracy / compliance check
• A safety standard is updated (every page citing it is revisited)

How a refresh is handled

1. Prioritise by how much traffic is at stake.
2. Protect winners — pages already performing well, and any paragraph being cited in an AI Overview, are left untouched.
3. Check the cluster — if several of your pages compete for the same topic, the system stops and consolidates the cluster around one strongest page, rather than refreshing them one by one.
4. Diagnose & fix only what's needed — wording, structure, schema, links, or intent.
5. Gate, package, publish, measure — same guardrails and before/after tracking as new content.

Modules

AI-Search (GEO)

Optimising not just for the classic search results, but for the AI answers — Google's AI Overviews, ChatGPT, Perplexity, and others.

What we do differently

• Answer-first writing — lead with a clear, liftable answer, then expand. AI engines reward this.
• Cover the whole question — map the full set of related questions a topic generates, so a page is comprehensive enough to be cited.
• Information gain — original, first-hand data you supply gives a page facts the AI hasn't seen elsewhere, which is one of the strongest signals for being cited.
• Be citation-worthy — clean structure and real, attributable facts and standards.
• Quality gate — each piece is scored on these dimensions and must pass before it ships.

Modules

The Skills

The system is made of focused, repeatable skills. Each does one job well; together they run the workflows. Status updates weekly through the trial.

Reference

Governance & Guardrails

The rules the system never breaks. They protect your brand, your customers' safety, and your compliance.

Reference

What We Track

The system keeps a complete, current picture of your US store. In plain terms, that's:

Reference

Tools & Connections

The system runs on a small set of best-in-class tools, all operated by Tasken on your behalf — there's no software for your team to learn, no keys or subscriptions to manage.

Reference

Markets

The current engagement is focused on one market, proven end-to-end.

The system is built market-aware from the ground up, so the same engine extends to additional stores without rebuilding it.

Reference

Glossary

Reference

FAQ

When will we see ranking improvements?

Rankings move after changes are live, generally over the following weeks — search engines need time to re-crawl and re-evaluate. We commit to the deliverables and report movement honestly as it happens.

Will anything change on our site automatically?

No. Every change is delivered as a package and published by your team. You stay in control of what goes live.

Can we feed in our own data or expertise?

Yes — and you should. Real installation data, test results, proprietary specs, field experience, and original photos make your pages genuinely different from competitors'. Supply whatever you have; the system weaves it in (after the usual review), which strengthens rankings, E-E-A-T, and the odds of being cited in AI answers.

How is this different from using an AI chatbot ourselves?

A chatbot is great for one-off questions. This is a persistent system with your product data, your safety standards, and cross-page consistency built in — applied to every piece of work, every week.

How do you keep fire-safety content accurate?

The exact code text is stored once and quoted verbatim, and a person reviews anything touching specifications or building regulations before it ships.

What do you need from us?

Publishing the packages in Magento, product images, expert review of safety specs, the occasional policy decision, and any unique data or first-hand information you'd like reflected in the content. We keep work that isn't blocked on you moving in parallel.