Article quality
How Citeca researches, structures and scores an article before it reaches your site.
An answer engine does not quote a page because it ranks. It quotes a page because one passage on it answers the question cleanly enough to lift. Every structural decision below exists to produce that passage.
What every article contains#
- A direct answer near the top. The question is answered in the first screen, in plain language, before any context or preamble.
- Key takeaways. A short summary block — the highest-value thing on the page for an engine assembling a response.
- Question-shaped headings. Sections titled the way people ask, so a retrieval system can match them.
- An FAQ block. Marked up so it is machine-readable, covering the follow-up questions the main body does not.
- Sourced claims. Figures and statements attributed rather than asserted.
- Internal links. To your related articles and to the product pages that matter, so authority flows somewhere useful.
- A hero image with real alt text. Described for what it shows.
Why the structure is so repetitive
Because it works. Retrieval systems reward predictable, extractable structure, and a page that hides its answer in paragraph nine does not get quoted no matter how well written paragraph nine is.
The article score#
Every draft is scored out of 100 before you see it, and the breakdown is shown beside the editor so you can tell what a change costs. It is a floor, not a grade — a draft that fails is rewritten rather than published.
| Check | What it wants | Why |
|---|---|---|
| Length | Enough words to cover the question properly | Thin pages get neither ranked nor quoted |
| Keyword in title | The prompt phrasing in the title | The strongest single relevance signal there is |
| Meta description | 150–160 characters | Controls the snippet, and gets truncated outside that range |
| Key takeaways | A summary block present | The passage most often lifted into an answer |
| FAQ | Several real follow-up questions | Directly matches how prompts are phrased |
| Hero image | An image with descriptive alt text | Needed for image results and accessibility |
Note
Editing is expected. The score recalculates live, so if you cut the FAQ you will see it drop and can decide whether you meant to.
Which models write what#
Different work suits different models, so Citeca does not use one for everything. Research and structure are planned by a reasoning model; drafting, formatting and markup are done by models chosen for accuracy and cost on that specific job. You never pick a model — that is our problem, not yours.
What Citeca will not do#
- Invent statistics. An unsourceable number is cut rather than guessed.
- Make claims you have not authorised. Certifications, guarantees and results come from your brand kit.
- Name competitors you have excluded. Your blocklist is enforced at writing time.
- Spin the same article ten ways. Near-duplicate pages compete with each other and devalue both.
- Publish a draft that failed scoring. It goes back to be rewritten.
Common questions
Can I rewrite an article completely?
Yes. The editor is a full editor and your version is what publishes. Republishing updates the post already on your site rather than adding a second one.
Will readers be able to tell it was AI-assisted?
Structure is deliberate and consistent, which reads as organised rather than robotic. The pieces that usually give it away — invented facts, filler paragraphs, no point of view — are what the scoring exists to catch.
Does Google penalise AI-assisted content?
Google penalises unhelpful content, whoever produced it. Its own guidance is explicit that how a page was made matters less than whether it demonstrably helps the person reading it. That is the bar the scoring is built around.
Can I supply the topics myself?
Yes. Add your own to the calendar with a specific instruction for each, and they are researched and written like any other.
Something missing or wrong here? Tell us and we will fix it.