Keywords and content

These tools cover the research-to-draft path: discover keywords, check whether interest is rising, group them into topic clusters, write an article from a topic in one step or from a researched brief, and record a human's approval of the draft. Nothing here publishes. Publishing is a separate, deliberate step.

The table above is generated from RankX AI itself rather than written by hand, as at 30 September 2026 (source b27dec96). When the product changes, this page changes with it.

The workflow, in order

graph LR
  A[research_keywords] --> B[save_keywords]
  B --> C[link_keyword_to_cluster]
  C --> D[create_content_brief]
  D --> E[get_content_brief]
  E --> F[generate_article]
  F --> G[get_content_item]
  G --> H[A human reviews and approves]
  H --> I[publish_content_item]
  W[write_article] --> R[get_article_run]
  R --> G

Each arrow is a decision point rather than an automatic hand-off, and the writing steps take minutes rather than seconds. write_article is the short path: one call from a topic to a draft, with the brief handled inside it.

Discovery and trends

research_keywords discovers related keywords for up to twenty seed terms with volume, cost per click, competition and trend. It spends only for a fresh discovery: repeating a recent search inside the cache window is free.

It needs no status tool, by design. If it reports that it is still running, call it again with the same seeds: the call is self-converging and safe to retry without spending twice.

get_keyword_trends is different in shape. It covers up to five keywords, and a fresh analysis is asynchronous: it returns a task id, and you poll get_keyword_trends_status until the data is ready. A repeat inside the cache window returns the data immediately and free. Related queries only come back for a single-keyword request.

Reading is always free, so polling costs nothing.

Topic clusters

list_topic_clusters reads a Website's clusters with their linked counts. A cluster is RankX AI's single source of truth for how keywords, tracked prompts and content group into a subject, so it is the join between the research side and the measurement side.

link_keyword_to_cluster attaches an already-tracked keyword to one. It is attach-only and idempotent: an existing link is never modified, so a retry cannot demote a link that already exists.

Clusters themselves are created in the product, deliberately: cluster structure is a decision about how the business describes itself, so the person who owns that decision makes it there.

Briefs

create_content_brief creates a brief for a primary keyword and starts its background research. Creating a brief spends credits, because that is when the research runs, and its own description carries the current price. Generating the article from the brief is a second, separately priced step, which is why write_article is the simpler choice when you have a topic and no brief.

Poll get_content_brief until the research completes. It returns the brief in full: primary and supporting keywords, the reader questions it should answer, the angle, the outline and the status.

Both brief tools are gated on the content-generation plan feature, so an account whose plan does not carry it will not see them listed at all.

Writing an article in one step

write_article writes a full article from a topic or keyword in one call: research, outline and writing run in the background for one price by length (short, standard or long), with no brief to create or approve first. Its own description carries the current price for each length, and it needs the spend scope, so the assistant confirms the cost with you before calling it.

It is a dispatch-and-poll pair. write_article returns at once with an id, and get_article_run reports the stage: researching, writing, or writing queued (another article on the same Website is being written first, so keep polling). Stop when it reports settled, at ready, needs edits (a saved draft that needs edits and will not move on by itself), failed or not started. Once the run reports an item id, get_content_item reads the draft.

The price is reserved when the run starts and charged once the article is written. Nothing is charged when the research finds no citable sources, or when the writing never starts. Like generate_article, it does not publish: the draft lands in Content Studio for a human to review.

Generating from a brief

generate_article writes the article from a brief you have already researched. It is a dispatch-and-poll pair too: it returns a content item id immediately, generation runs in the background over a few minutes, and you poll get_content_item while the status is generating or in quality review. Ready for review or changes requested (a saved draft that needs edits) means there is a draft to read; archived, cancelled or failed means the run stopped.

Four properties worth knowing before you call it:

It needs a brief with completed research. A brief whose research has not finished is refused before any credits are reserved, so a premature call costs nothing.

It approves a draft brief as part of the step. A confirmed generate is taken as approval of the brief it is generating from.

Its price is tiered by length, and the tool's own description carries the current figure. This documentation does not print prices, because rates are versioned and republishable.

It does not publish. The draft lands in Content Studio for a human to review. Publishing to a connected site is a separate group of tools, under a different scope.

Approving a draft

approve_content_item records the human sign-off an article needs before it can be published. It is free and changes nothing on the website: only a draft that is ready for review, or has had changes requested, can be approved, and an article carrying a quote from a named person is refused until that person has agreed to their words. Show the person the article with get_content_item first and approve only when they say so, because approval is the decision, not a formality.

publish_content_item is what then puts it on the site, and it refuses an article nobody has approved.

Reading a draft back

list_content_items returns items with status, target keyword and quality scores. get_content_item returns one in full: the body as plain text, meta title and description, target keywords, quality and GEO scores, and counts of fact-checked claims and internal links.

A null score means not yet computed, not zero. A null body means the article has not been drafted yet, which is what you will see while a generation is still running.

Example prompts

"Research keywords around 'ai visibility tracking' and show me the ten with the best volume-to-competition ratio. Do not save anything yet."

"Save the five keywords I just picked for tracking, then link them to my 'AI search' topic cluster."

"Check the search trend for 'generative engine optimisation'. If it is still processing, wait and check again."

"Create a content brief for 'how to appear in AI answers', then tell me when the research is done and summarise the outline."

"Generate an article from brief X. Confirm the cost with me first, then poll until it is ready and show me the first three paragraphs."

"Write a standard-length article on 'how AI assistants choose sources'. Tell me the price first, then poll until it settles and show me the outline."

Where to go next

• Content briefs for the workflow in the product.
• Publishing to WordPress for what happens after review.
• Credits and metering for how the spend tools are charged.

Last updated 1 October 2026