A brand-new ed-tech domain with no history and no authority. An n8n-driven content engine built to earn indexation, rankings and AI-search visibility inside the first eight weeks, without publishing a single page of thin content.
New domains start with no crawl history, no trust signals and no ranking footprint. Writing a few articles a week by hand means months of near-zero visibility. Automating naively means thin, repetitive pages that a spam update will erase. The brief was to be fast and defensible at the same time.
The principle behind the build: automate the process, not the judgment. Every decision a strategist would make (which topic, how hard, how similar to what already exists, whether the page is good enough to publish) was turned into an explicit rule or gate inside the workflow. Speed came from removing manual hand-offs, not from lowering the bar.
The engine is a single n8n workflow of roughly a hundred nodes, organised into three layers. Each layer owns one question: what to write, whether it is fit to publish, and how it reaches the reader and the machine reader.
The sheet is the source of truth: topic, keyword tier, segment, status. The workflow reads from it and writes status back, so nothing is tracked in anyone's head.
Seeds are vetted first, then expanded in controlled steps. Each new topic is tied to a tier so effort follows opportunity.
Topics are staged by how hard they are to rank for. A brand-new domain starts on winnable terms and earns the right to attempt harder ones.
Articles are produced across eight segments, each with its own structure, so the site reads as a library rather than one repeated template.
A Jaccard-similarity check compares every candidate against existing titles and topics and rejects overlaps before a draft is generated, which prevents cannibalisation.
Drafts are scored against defined criteria. Anything under the bar is regenerated or held back instead of published.
Titles, descriptions and slugs are checked for length, uniqueness and intent match before the post is created.
Generation runs through OpenRouter with Claude as the writing model, so models can be swapped without rebuilding the workflow.
Posts are created through the WordPress API with the right category, metadata and conversion banner attached, with no copy-paste step.
Runs start from a Telegram message, and the workflow reports back there. Operating the engine does not require opening n8n.
Search is now split between ten blue links and AI-generated answers. The engine writes for both: definition-first openings that an answer engine can lift cleanly, question-style headings that match how people phrase queries, and structured sections with FAQ blocks. The same page works as a snippet candidate, a citation candidate and a normal result.
The domain showed no organic footprint through February 2026, picked up its first rankings in March, and kept climbing as new clusters were indexed. It now holds the number-one position for a high-volume head term plus a group of closely related definition queries, and top-4 positions on interview-question topics, all from content the engine produced.
Source: Ahrefs estimates as of October 2026. These are third-party estimates, not Google Search Console data, and Ahrefs tends to under-report long-tail traffic. The dip in June is shown as reported. Search Console figures can be shared on request.
Start on winnable terms. Difficulty staging lets a zero-authority site collect early rankings instead of publishing into positions it cannot reach yet.
Put the gates in code. De-duplication, scoring and metadata checks are what make automated output safe to publish at speed.
Write for the machine reader too. Answer-first structure costs nothing and positions the content for AI search as well as Google.
Keep the human on the controls. The strategist sets tiers, thresholds and segments. The workflow only executes them.
Want an engine like this for a new domain or a stalled content programme? Talk to SEOBiopse or reach out on LinkedIn.