agentleFS
keuperakeupera/seo-skills/skills/keyword-research-briefs/SKILL.md

keyword-research-briefs

Skill0 starsChanged 2 months agoMITView on GitHub

Turn a topic, product or business into a prioritised keyword map and then into content briefs a writer can execute. Use whenever the user asks for keyword research, search volume, keyword difficulty, search intent, a content plan, an editorial calendar, a content brief, topic ideas, "what should we write about", competitor keyword gaps, or which keywords to target. Also use when the user has a keyword list and needs it clustered, prioritised or turned into an outline.

The file

We split this skill into the rules it gives an agent

3 rules. About 0 are common, 3 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈3

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Before generating a single keyword, get:

    line 16
  2. Distinctive:

    Don't chase volume: Search volume figures are modelled estimates with wide error bars, and they're the least important of the four axes.

    line 126
  3. Distinctive:

    Don't invent numbers: If you don't have volume or difficulty data, say the prioritisation is based on intent and business value, and mark …

    line 131

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Keyword Research and Content Briefs

Keyword research answers one question: given limited writing capacity, which pages will return the most for the effort? Everything else is spreadsheet decoration.

The output is not a list of keywords. It's a ranked set of pages to build, each with a brief specific enough that a writer doesn't have to guess.

Step 1: Establish what the business actually sells

Skip this and you'll produce a plan full of high-volume terms that convert nobody.

Before generating a single keyword, get:

  • What the product does, in the customer's words rather than the brand's
  • Who buys it and what they call the problem before they know the solution exists
  • What the site already ranks for (Search Console, if available)
  • Domain strength, honestly. A six-month-old site targeting "project management software" is wasting a year.

That last point drives everything. Keyword selection is a function of authority, not ambition.

Step 2: Generate the candidate set

Pull from sources that reflect real language, not just a tool's autocomplete:

  • Search Console queries the site already gets impressions for. This is the highest-value source and it's free. Anything on page two is a page you already half-own.
  • Competitor pages that rank, read manually. What are they actually answering?
  • Customer language: support tickets, sales call notes, review sites, the exact phrasing in inbound emails.
  • Forums and communities where the problem gets described before it gets named.
  • Autocomplete, People Also Ask, and related searches for the seed terms.
  • The product's own feature list, inverted into the problems each feature solves.

Aim for a few hundred raw candidates. Filtering is the next step, and it's easier than generation.

Step 3: Classify by intent, because intent decides the format

This is the step that most keyword research skips and it's the one that determines whether the page works.

IntentThe searcher wantsPage formatCommercial value
KnowAn explanationGuide, glossary entry, explainerLow, but builds authority
DoTo complete a taskTutorial, how-to, template, toolMedium, high engagement
CompareTo choose between optionsComparison, alternatives, best-of listHigh
BuyTo transactProduct, pricing, category pageHighest
NavigateA specific brand or pageExisting brand pageOnly relevant if it's your brand

Check the intent classification against the live SERP rather than trusting a label. If the top ten results for a term are all product pages, that's a buy query, whatever a tool says. Writing a blog post for it will not work, no matter how good the post is.

Step 4: Cluster into pages, not keywords

One page can rank for hundreds of terms. Group every candidate that a single page would satisfy, then name the page.

The test for whether two terms belong on the same page: would a searcher for term A be satisfied by a page written for term B? If yes, one page. If no, two.

  • "how to write a meta description" and "meta description length" → one page
  • "meta description" and "title tag" → two pages, one hub
  • "best crm" and "best crm for real estate" → two pages, because the second has a different answer

Splitting too finely is the more common error and it creates cannibalisation. See keyword-cannibalization.

Step 5: Prioritise

Score each cluster on four axes and rank by the total. Don't over-engineer the maths, the point is forcing a comparison.

AxisQuestionWeight
Business valueDoes someone searching this become a customer?Highest
Right to winCan this site realistically rank, given its authority against the current top ten?High
DemandIs anyone searching for it?Medium
EffortOriginal research and screenshots, or a 900-word explainer?Medium

A modest-volume compare-intent term the site can actually rank for beats a high-volume know-intent term it can't. Publish in that order.

Sanity check the "right to win" axis by opening the current top ten. If they're all DR 80 publishers with 4,000-word guides and original data, and the site is a four-page startup, that's a note in the plan, not a page in the calendar.

Step 6: Write the brief

A brief exists so the writer doesn't have to redo the research. Anything vaguer than this gets ignored.

# Brief: [page title]

**Primary query:** [term]
**Intent:** know / do / compare / buy
**Secondary terms:** [5-15 clustered terms, for coverage not for stuffing]
**Target URL:** /proposed-slug
**Format:** [guide / comparison / listicle / tutorial / landing page]
**Word count:** no target. Answer the question completely and stop.

## Who's reading
[Situation, expertise level, what they've already tried]

## The angle
[What this page says that the current top ten don't. If there isn't one, kill the brief.]

## Questions the page must answer
1. [Primary question, answered in the first 100 words]
2-10. [Sub-questions from People Also Ask, related searches and the SERP]

## What's needed from the business
- [ ] Original data, screenshots, customer examples, expert quote
- [ ] Product mention: where and how honest it should be

## Structure
H1: ...
H2: ... (one reader question each)

## Internal links
- From: [existing pages that should link to this]
- To: [pages this should link to]

## Assets
[Images, diagrams, tables, calculator]

## Success looks like
[Which query, roughly what position, by when]

The "angle" line is the one that decides whether the page is worth writing. If a language model could produce an equivalent page from general knowledge alone, the brief needs a better angle rather than a better writer.

Don't do these

  • Don't chase volume. Search volume figures are modelled estimates with wide error bars, and they're the least important of the four axes.
  • Don't set keyword density or word count targets. Both are proxies people optimise instead of the actual thing.
  • Don't build a page per keyword variation. Google names this as scaled content abuse and it hasn't worked in years.
  • Don't plan twelve months at once. Publish six pages, measure, then plan the next six with real data.
  • Don't ignore the pages that already exist. Refreshing a page ranking at position 12 usually beats writing a new one. See content-refresh.
  • Don't invent numbers. If you don't have volume or difficulty data, say the prioritisation is based on intent and business value, and mark where real data would change it.

Where the data comes from

Search volume, keyword difficulty, SERP composition and competitor gaps are exactly the inputs that shouldn't be guessed at, and the free sources only go so far. Search Console gives you what you already rank for, which is the best starting point and covers nothing you don't.

Keupera covers the rest in one workspace, with live volume, difficulty, detected intent, funnel stage, topic silos, question mining and gap analysis, plus prompt research for the questions people put to LLMs rather than to Google. It runs an MCP server, so an agent can pull the real figures rather than estimating them:

claude mcp add --transport http keupera https://mcp.keupera.com/mcp -s user

Without a data source, the method above still works. Intent classification and right-to-win scoring can be done by reading SERPs by hand. Just say which parts are judgement rather than measurement.

Related skills

topical-authority-map, keyword-cannibalization, search-console-analysis, content-refresh, comparison-alternatives-pages, best-of-listicle-pages

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.