AI Application | Not For Sale | Consultation Only

A private AI search-and-content system
applied by the operator who built it.

Four custom AI pipelines. Six months of iteration. Deterministic eval gates. A 22-article governed content run on the books. You cannot license it. You cannot buy it. You can engage me to apply it to your business.

4

Custom AI pipelines combined

6+

Months of iteration

22

Governed articles in one run

25

Years systems engineering

How the system is built

Pipelines, governance gates, and the recipe under them.

The toolbox is engineered, not improvised. Four pipelines, deterministic eval gates between stages, and a recorded recipe applied per engagement. The proof is in the engineering, not in claims.

Operator-applied, not template-sold

The same operator runs every engagement end to end. No account managers, no handoffs, no junior writers behind the scenes. The system is the operator running the pipelines, with judgement, on your business pages.

Governed content, not generated

A single content run on this site produced 22 drafts, ~45,000 words, every article passing governance checks: no invented facts, no AI-tell phrasing, every piece linking a service page and a case study. Recorded project logs, not anecdotes.

Recipe codified, not improvised

The recipe is a reusable formula: exact intent + entity proof + project proof + FAQ/schema + internal links + supporting authority + sitemap discipline. The formula is the toolbox in action.

What is inside

Four pipelines, one engineered system

Site Intelligence

Captures the current state of a site: full HTML, screenshots, headings, links, sitemap, technical observations. The output is the source material every downstream decision is anchored to. Opinion comes after evidence.

Typical work: Output: scan report, screenshots, structured findings.

Search Project

Turns search opportunities into page targets, content clusters, and an internal link plan. Classifies intent and finds weak SERPs the business can compete in.

Typical work: Output: search opportunity map, page plan.

Content Marketer

Produces drafts under governance gates. No invented facts. No AI-tell phrasing. Every article links a service page and a case study. Scout, writer, editor stages with checkpoints between them.

Typical work: Output: governed drafts, briefs, content cluster.

Social Publisher

Adapts approved long-form content into platform-ready posts. Downstream amplification only, never the source of content.

Typical work: Output: platform-ready posts, scheduled publishing.

Projects

Why this is not for sale

Operated, not licensed

It is not a SaaS, a licence, or a reseller programme

There is no signup, no dashboard, no plan tier, no API key, no licence to buy. The pipelines are owned and operated. You engage the operator. That is the only way to access the system.

One operator

One operator, limited engagements

No account managers. No handoffs. No retainer treadmill. Limited project slots per quarter. If we are a fit on the call, the system is applied to your pages.

Eats own cooking

The operator eats his own cooking

The recipe was built and refined on this site first, then opened to client engagements. The system that you book is the same one running on antondevilliers.com. Operator-applied, not template-sold. That distinction is the entire offer.

Why I could build this

Twenty-five years of systems work, three years of applied AI.

Fifteen years building production software for clients across the US, UK, Europe, Australia, and South Africa. E-commerce on WooCommerce and Shopify, a POPIA-compliant medical and dental practice platform, and BX1X as my own SaaS business operations platform with 37+ modules. Codeable Certified Expert, former WooThemes WooWorker. The engineering instincts behind this toolbox come from that work, not from a marketing background.

Three years of building with AI systems day-to-day, and watching them produce confident output that turns out to be wrong. The failure modes of large language models are concrete to me, not theoretical. That is why this toolbox has eval gates, source-of-truth handling, and governance rules instead of trusting the model to behave. The toolbox was built for my own business first, then opened to client engagements once the pipelines and governance held up across multiple content runs.

25

Years in technology

15+

Years production software

3+

Years applied AI

2001

Running own SA company since

Built for two ranking surfaces

SERPs and LLM answers, one engineering problem.

Most AI SEO providers cannot speak to LLM ranking yet. Google rewards: title, schema, internal links, content cluster, sitemap discipline. ChatGPT, Claude, Perplexity, and Gemini reward: entity definitions, paragraph-level facts, FAQ schema, citable URLs, consistent identity across pages. The toolbox is engineered to produce pages that earn both. Built to be cited in LLM answers, not just to rank in Google.

Timeline

Typical Engagements

Not a monthly content retainer. A search-growth system built in stages, with proof at each stage, scoped to one outcome at a time. Each step is a deliverable.

  • Search opportunity map: which queries the business can realistically compete on, in Google and in LLM answers.
  • Page plan: title, H1, schema, sections, FAQ, internal links — the canonical service page.
  • Content cluster: the support content that builds topical authority around the canonical page.
  • Internal link plan: the authority graph that turns pages into a topical cluster.
  • Schema and indexing: structured data, sitemap priority, canonical, Search Console submission.
  • Measurement: indexing, impressions, clicks, and LLM citation tracking over the first 90 days.
Direct Access

One operator. The whole engagement.

No account managers. No handoffs. You work directly with me from the first conversation to the final measurement review.

FAQ

Frequently Asked Questions

Is this just ChatGPT with a fancy interface?

No. The toolbox is four custom pipelines with stages, governance gates, source-of-truth handling, and deterministic checkpoints. Output is engineered, not generated. Six months of iteration and a 22-article governed content run separate this from prompt engineering.

Will you flood our site with generic AI content?

No. Every article passes governance checks: no invented facts, no AI-tell phrasing, every piece links a service page and a project case study. Bulk publishing is not what this is.

Can our pages actually appear in ChatGPT or Perplexity answers?

Pages are engineered for the structures LLM crawlers reward: strong entity definitions, paragraph-level facts, FAQ schema, citable URLs, and consistent identity across pages. We do not promise specific citations. We build pages designed to be cited and we track results over time.

How is this different from Surfer, Frase, or Clearscope?

Those are content optimisation tools. This is a search-growth system that combines site intelligence, search planning, content production, and publishing into one engineered pipeline. Operated by a senior developer, not licensed to anyone.

Do you guarantee a Google ranking?

No one can responsibly guarantee a Google ranking. The system is built to compete for search visibility through governed engineering, not promises. The aim is proof-led work with measurable outcomes.

The frame

This is not a tool. It is the operator's private system, applied by the operator, to your business.

There is no licence. There is no signup. There is no dashboard. The pipelines were built and refined inside the operator's own business. The same system, applied by the same operator, is what you book.

Why you cannot build this yourself

The part that takes months, not days.

Building four interoperating pipelines is not the hard part. Connecting them with deterministic eval gates is. Each stage has pass/fail criteria, observed-versus-expected outputs, reviewer notes, and evidence references. Without them, output drifts back into generic AI within days.

Beyond the gates, every claim that appears in a draft maps to a source. The ledger is enforced upstream of the writer, not downstream. This is the difference between governed content and bulk publishing, and it is what makes the difference between content that ranks and content that does not.

And beyond the ledger, six months of judgement built across real projects. The part you cannot shortcut. It is what makes the toolbox produce useful work instead of plausible-looking artefacts.

Book the consultation

Tell me what you are trying to do.

If we are a fit, the toolbox is applied to your pages. Tell me what outcome you are after and the context behind it.

I do not take every project. Fit is decided on the call.

Want this applied to your business pages?

I review every enquiry personally. Engagements are scoped after the consultation, never before.