Content engine
Not a chatbot with a publish button.
Every article runs a multi-stage pipeline before a person ever sees it: written against real keyword data — with humanization and the brand voice built into the draft — then fact-checked and scored against SEO checks. Anything questionable stops and waits for a human.
The pipeline
One prompt can produce an article. It can't produce a good one reliably. The engine splits the work across specialized stages, each with one job:
- Write — drafted against a volume-backed focus keyphrase, the business context, and the pillar's angle registry, with answer-first formatting for featured snippets; humanization rules (no AI tells, no stiff phrasing) and the brand voice defined during setup are built into the draft itself
- Statistics with receipts — a live web-research pass gathers real numbers for your industry, chases each one past the retellings to its original source, and verifies it before an article can cite it. Competitor sites are never cited.
- Fact-check — claims verified; the engine is instructed to flag uncertainty rather than invent specifics
- SEO pass — keyphrase placement, headings, internal links, and metadata checked deterministically, not by vibes
- Metadata written like prose, sized like Google: SEO titles fit the ~60 characters Google actually displays, meta descriptions land at 120–155 and always end on a complete sentence — validated and rewritten when wrong, never machine-patched
- Social posts — matching posts drafted for every connected platform, rotated across post archetypes so a feed never looks templated
A quality gate that doesn't trust the AI
After the pipeline, a deterministic scanner — plain code, not a model grading itself — checks every article for placeholders, repetition, keyphrase coverage, word-count bands, and more. Clean pieces move on automatically. Real problems stop the piece in the Review column with the reasons listed; advisory notes ride along without blocking, so the queue holds what actually needs a human. Nothing questionable publishes itself.
Grounded in the business, not the internet average
Setup teaches the engine the business once: a scan of the website and sitemap, the services and service areas, personas and ideal customers, and the brand voice. Every article draws on that context — which is why the output reads like the business wrote it, not like a content farm found the industry on Wikipedia.
- Website and sitemap scan maps services, service areas, and the pages that make money
- Personas and ideal customer profiles shape angle selection and vocabulary
- Structured data on every article: Article, FAQ, and LocalBusiness schema where each applies
- Original imagery, not stock: hero images are custom diagrams drawn from the article's own facts, or a branded card — stock photography is the last resort, not the default
- Editing stays human: a rich-text editor with AI expand/shorten/rephrase on selection, when you want to touch a draft
- Every change keeps its history: each edit to an article — yours or the engine's — saves the version it replaced, with a diff and one-click restore
See what it would target for your business.
Tell us your website and your trade — we'll show you the keyword picture and how the engine would fill it.