A writer with a keyword and no brief produces a page that ranks for nothing. This playbook turns a raw keyword list into a full content brief — search intent, outline, entities to cover, gaps in the pages that currently rank, and internal-link targets — so whoever writes the piece starts with a map instead of a hunch.
Workflow area: Marketing · Tier: Core · Hands-on time: ~30 minutes
The brief is the highest-leverage document in content marketing because it is written once and read by everyone downstream: the writer, the editor, the designer choosing images, and future-you deciding whether the piece worked. AI is genuinely good at the brief's grunt work — clustering keywords, inferring intent, structuring outlines — as long as you feed it real search-page evidence instead of letting it guess what ranks.
Cluster the keyword list first. Paste your raw list and let the model group it — one brief per cluster, not per keyword:
Here is a raw keyword list from my keyword tool. Group these into
clusters where one article could plausibly rank for the whole cluster.
For each cluster give: a cluster name, the primary keyword (highest
intent), secondary keywords, and the likely search intent
(informational / comparison / transactional / navigational).
Keywords:
[PASTE LIST — one per line, include volume if you have it]
Rules: do not invent keywords that are not in my list. If two
keywords have clearly different intents, put them in different
clusters even if the words look similar.
Gather real evidence for the cluster you're briefing. Search the primary keyword yourself. Copy the titles and H2s of the top 3–5 ranking pages (two minutes with each page's headings, or a scrape from your SEO tool). This step is what separates a grounded brief from a hallucinated one.
Generate the brief from the evidence. One prompt, everything in:
Create an SEO content brief for an article targeting this cluster.
Primary keyword: [KEYWORD]
Secondary keywords: [LIST]
Search intent: [FROM STEP 1]
Our site: [ONE LINE — who we are, who reads us]
Audience: [e.g. "HR managers at 50–500 person companies in India"]
Top-ranking pages and their headings:
1. [TITLE] — [H2s]
2. [TITLE] — [H2s]
3. [TITLE] — [H2s]
Produce:
1. Working title (under 60 characters) + 2 alternatives
2. Meta description (under 155 characters)
3. Search intent statement — what the searcher is actually trying
to do, in one sentence
4. Recommended outline: H2s and H3s, with a one-line note per H2 on
what it must cover
5. Entities and terms to include (people, tools, standards, concepts
the topic requires)
6. Content gaps — what NONE of the top pages cover well, based only
on the headings I gave you
7. Questions to answer (People-Also-Ask style)
8. Suggested word-count range and format (guide / listicle /
comparison)
9. Internal-link targets: from this list of our existing pages,
which 3 should this article link to and why: [PASTE PAGE LIST]
Rules: base the gap analysis only on the evidence I pasted — say
"based on headings provided" where you are inferring. Do not invent
statistics or claim what competitors' full text says.
Sanity-check the brief against the live results. Good output for step 6 looks like this:
Content gaps (based on headings provided):
- None of the three pages address pricing for the Indian market —
all examples are in USD.
- No page includes a comparison table; all are prose-only guides.
- Only page 2 mentions compliance, and only in passing — a
dedicated H2 on DPDP Act implications would be differentiating.
If a "gap" is actually covered by a ranking page (the model only saw headings, not body text), strike it. If the outline mirrors competitor pages so closely it adds nothing, push back: "The outline is a clone of page 1. Restructure it around the gaps."
Package for the writer. Paste the final brief into a Google Doc, add your own two lines at the top — the angle only you know (a customer quote, a data point, an opinion) — and assign it. The human angle line is what stops ten companies with the same AI from publishing the same article.
Keyword lists and competitor headings are public data — this playbook is safe for consumer AI tools. Keep unreleased product details, embargoed launches, and internal traffic numbers out of the brief prompt, or run it in your approved enterprise tenant.
The model structures the brief; the evidence makes it true. Two minutes of copying real search-result headings is the difference between a brief grounded in what actually ranks and a confident guess wearing the same format.