A language model builds the outline and the first draft against a brief from your keyword map; an editor checks the facts against your sources and adds what competitors don't have. Volume follows the structure of the site, not a quota.
Content is usually the bottleneck: the structure is planned and there's nobody to fill it. We use a language model where it genuinely saves time — the outline from a brief, a first draft, repetitive category descriptions, meta tag variants across hundreds of pages. Then comes the part a model cannot do: checking facts against your sources, adding your prices and terms, your limitations, the mistakes customers typically make, examples from practice. Search engines don't penalise the use of a model; they penalise pages with no value — a retelling of what already ranks. So our rule is simple: if a piece adds nothing to what a reader can already find in the results, we don't publish it and we tell you why. In health, finance and legal work a draft is scaffolding only — the facts and the wording come from a qualified person.
Copywriters produce 5-10 articles/month. Covering 500 keywords takes 1-2 years — competitors get there first.
Covering a niche means dozens of pages. At market writing rates the budget runs into five figures, and part of it never earns anything back.
Product pages ship without their own title and description. Search engines improvise, and the snippet loses to competitors before anyone clicks.
A literal translation doesn't work: people phrase queries differently in each language. What's needed is localisation to local demand, not a word-for-word copy.
To appear in assistant answers a page needs short, exact answers to real questions. Doing that by hand across a whole site takes months.
Blind generation without processing — Google flags as AI spam and downranks or deindexes.
Expert-level content, 1,500-4,000 words, structured for readability and search intent. LLM writes, editor verifies facts and adds expertise.
Unique title and description for 100-10,000 pages. We consider search intent, relevant keywords, length limits.
8-15 Q&A pairs for each commercial page. Format optimized for FAQPage schema and AI Overview capture.
Unique descriptions for 100-5,000 categories with niche-specific context and relevant keywords.
Bulk update for 3+ year old articles: fresh statistics, current formats, FAQ additions, updated examples.
Translation preserving SEO optimization: adapting keywords to target language, local examples, cultural nuances.
Post-processing to defeat AI detectors, adding personal turns of phrase, specific examples, case studies. Passes GPTZero and Originality.ai.
Every piece is checked before publication: facts against your sources, originality, readability, and whether it actually answers the query better than the pages already ranking
Content is part of the SEO retainer: we set the monthly volume during the audit, based on your structure and the competition. If you need more pieces than the tier covers, the pricing page lists the add-on.
We take it standalone when another vendor runs your SEO and only the content side is missing — scoped by volume and by how demanding the subject matter is.
In regulated fields — health, finance, legal — timelines are longer: a qualified reviewer is mandatory.
100 articles in 10-14 days instead of 6 months. You capture most niches before competitors even start.
Not 'just text' — targeted for specific keywords, intent, LSI. Indexes in the first week.
Every article verified by a subject-matter editor with domain credentials. Google distinguishes this from 'raw' AI content.
Prices, terms and regulations come from your documents. Anything the model recalled on its own is cut before editing, not after publication.
Collect 200-1,000 target keywords, cluster via AI, prepare briefs for each article with keywords, LSI, structure.
LLMs (GPT-4 + Claude) generate drafts against strict briefs. 20-50 articles per day.
SEO editor verifies facts, adds expertise, references, examples. Passes quality checklist.
Final review of keyword density, LSI, H1-H6 structure. Unique title/description generation.
Upload to CMS via API. Monitor indexing and rankings. 30-day traffic report.
Every piece maps to a cluster: which query, what the reader wants, what the answer must contain, and what the pages currently ranking are missing.
A language model assembles the outline and the first draft against the brief. That saves time on structure; it does not replace expertise.
Prices, timelines, terms and regulations are checked against your sources. Anything the model "remembered" on its own is removed.
The part nobody else has: your terms, your limitations, the mistakes customers typically make, real examples from practice.
Cutting filler, checking the logic, adding tables and direct answers that assistants can quote.
Author, update date, schema, internal links. After two or three months we look at what the piece brought in and improve it.
Penalties don't come from using a language model. They come from pages with no value: a retelling of what already ranks, duplicates, copy that doesn't match the intent of the query. In practice a draft that has been fact-checked and edited by someone who knows the subject ranks alongside writing produced from scratch. So our line isn't drawn at the tool but at the result: if a piece adds nothing to what the reader can already find in the results, we don't publish it and we tell you so.
No article publishes without human editor review. Otherwise it's AI spam.
Not '100 articles around a topic' — but '100 articles for 100 specific queries with their search intents'.
LLMs hallucinate — editor verifies statistics, dates, quotes, links. False facts destroy trust.
We train LLMs on 5-10 of your reference articles. Result is in your brand's voice, not default ChatGPT.
They penalise pages with no value: a retelling of what already ranks, duplicates, copy that doesn't match the intent of the query. A draft that has been fact-checked and edited by someone who knows the subject ranks alongside writing produced from scratch. The risk appears where generation is run at scale without an editor — and those are exactly the pages that end up being deleted later.
We are at the preparation stage, you are at sign-off. Every figure, price, lead time and condition comes from your sources rather than from the model's memory: anything it recalled on its own gets removed. In demanding subject areas your specialist reviews the piece before publication. That isn't a formality — a confident error about a price or a regulation costs more than the article did.
As many as the structure of the site needs, not as many as the model can generate. Usually that's a handful of pieces plus work on existing pages, which are often worth improving rather than replacing. We set the volume during the audit: in a narrow niche a dozen strong pages cover the demand, while a large catalogue is worked through section by section.
Current pricing and terms, a specialist who can answer a few questions a month, and access to the questions customers actually ask by email, chat or phone. That last one is the most valuable source of topics: it shows what people genuinely don't understand, rather than what seems important internally.
Reviews and customer stories — never. That's what costs a site its credibility outright, and it's easier to detect than people assume. Case studies we write only for real projects, with figures from your own analytics and the period stated. If the numbers can't be substantiated, we describe the work without percentage claims.
Yes, and it shortens the work: if you already have clusters and a page map we start straight on the pieces. We need the page URL, the primary query and what the reader wants to learn. If the keyword research is old, it's worth reviewing first — part of it is usually out of date.
Free 30-minute consultation