Content
AI that researches, cites its sources and prepares a WordPress draft: Blog Writer
Blog Writer researches a topic or keyword, writes an original post that cites the sources it read, and returns the title, slug, meta title, meta description, excerpt, HTML body, FAQ and image ideas together. The fields map directly onto WordPress Create post; the post is saved as a draft and you publish it.
- As a workflow node
agent.blogWriter- Default model
- Claude Sonnet 5
- Economy model
- Claude Haiku 4.5
- Tools
- 4
- Report
- Blog post
What the agent does
Give it a topic and it writes that. With neither topic nor keyword it picks one itself: from queries ranking 5–20 with many impressions when Search Console is connected, otherwise from the topic pillars of your strategy. When the previous step passes in the site's posts (WordPress List posts), it does not repeat a topic and links to relevant posts.
It researches with web search, page reading and your knowledge base. Every number, date and quote comes from a source it read, is linked in the text and listed in the sources; it never invents sources, statistics or experts. It never copies another page's sentences or outline. No health promises, no unproven “best” claims and no attacks on competitors.
The rules are checked in code: meta title at most 60 and meta description at most 155 characters, at least two H2 sections, the requested length and an FAQ rendered in the body. The body HTML is cleaned: scripts, styles, images and unsafe links are removed and an H1 becomes an H2. In a workflow the post flows as one item; the ready-made template Weekly blog post → WordPress draft prepares a draft every Monday.
Inputs / Outputs
Inputs
- Topic
- What the post is about. Fill in the topic or the keyword; with both empty the agent picks one from your strategy and Search Console.
- Main keyword
- The search term the post should rank for; derived from the topic when empty.
- Outline
- Optional H2 headings in the order you want. Empty means the outline follows the search intent.
- Word count
- 600–2500 words, default 1200. Comprehensive guides 1800–2500.
- Audience
- Who reads it and what they already know. Empty means the project's target audience.
- Language
- Language of the post as a two-letter code (tr, en …). Turkish by default.
- Internal links
- Product, category or service pages the post should link to; each is used once with a descriptive anchor.
- Add an FAQ section
- When on, the post ends with an FAQ of 3–6 questions, also returned as a separate field.
Outputs
- title
- The post title (WordPress title, H1).
- slug
- URL part built from the focus keyword, with Turkish characters transliterated.
- metaTitle
- SEO title of at most 60 characters, keyword first.
- metaDescription
- Meta description of at most 155 characters.
- excerpt
- A 1–2 sentence summary for archives and the WordPress excerpt.
- bodyHtml
- Cleaned HTML body with H2/H3 headings and source links; the FAQ at the end when requested.
- headings
- The H2 and H3 headings of the body.
- faq
- Questions and answers, as a separate field for FAQ structured data.
- sources
- Title and link of every source used in the post.
- suggestedImages
- Image ideas with alt text for the cover and sections.
- tags
- 3–8 WordPress tags.
- wordCount
- Word count of the body.
Saves a report to the Reports page on every run: Blog post.
Tools
- Web search
web_search - Read a web page
fetch_url - Knowledge base search
knowledge_search - Search Console data
get_search_console
Integrations it uses
Example tasks
- “Write a 1200-word guide on choosing a coffee grinder for filter coffee for home baristas; link to our grinders page.”
- “Write a post with an FAQ that answers 'how much coffee per cup for V60' and cite your sources.”
- “Pick this week's topic from our Search Console queries and write a post our site does not have yet.”
- “Prepare a 1500-word post for office managers on 7 criteria for choosing a coffee supplier.”
Tips for good results
- Connect Serper or Tavily for researched posts; without them the agent writes from the knowledge base and general knowledge, without statistics.
- Put WordPress List posts before the agent; it will not repeat existing posts and links to them.
- Copy the meta title and meta description into Yoast or Rank Math; use your real photos for the suggested images.
- Open the sources and check the facts before publishing; the post waits in WordPress as a draft.
Model advice
Default model
Claude Sonnet 5
Per 1M tokens: $2 input · $10 output
Economy model
Claude Haiku 4.5
Per 1M tokens: $1 input · $5 output
The default balanced model is good at turning research into an original, fluent post and matching facts to sources correctly. For short posts and simple topics the economy model is enough; stay on the balanced model for comprehensive guides over 2000 words.
Templates that use this agent
Frequently asked questions
Does it publish the post on WordPress?
Not in the ready-made template: after your approval the post is saved to WordPress as a draft and you publish it. The agent itself does not write to WordPress; the WordPress step of the workflow does.
Where do the sources come from?
From pages it found with web search and actually opened, and from your knowledge base. Every source it uses is linked in the text and listed; it never lists a source it did not read.
Could the post be copied from another site?
No. The agent reads ranking pages to understand the intent but never copies their sentences or outline; it uses at most two short quotes, in quotation marks and attributed.
Can it choose the topic itself?
Yes. With topic and keyword empty it picks a topic your site does not cover yet from your Search Console queries or the topic pillars of your strategy.
Get started with OrqLabs
Build your first workflow today
Start on the free plan: connect your accounts, pick a template and stay in control with approval steps. No credit card required.