Claude prompt for a SERP-informed content brief.

Paste a target query and the top 10 search results. Get back a brief that matches table stakes, beats on identified content gaps, and refuses to pad. Includes a "what NOT to write" section listing the moves to skip.

The prompt (copy and paste)

Target query: {{QUERY}}. I'll paste the top results as {{SERP_CONTENT}}.

Analyze:
- Dominant intent and sub-intents present
- Consensus structure (sections most results share)
- Content gaps no one covers well
- Expertise signals we'd need to match
- Format conventions (tables, calculators, images)

Then write a brief that: matches table stakes, beats on the identified gaps, and refuses to pad. Include a "what NOT to write" section listing common moves we'll skip.

What Claude actually produces (real sample)

Sample run against the query "how to set up claude projects for a marketing team" with 10 results pasted as input. Real first-pass Claude output, not cherry-picked.

**Target query:** "how to set up claude projects for a marketing team"

**Dominant intent:** Operator-led tactical setup. Sub-intents: (1) "what is a Claude Project," (2) "how Projects differ from regular chat," (3) "what to put in the system prompt for marketing teams."

**Consensus structure across top 10 (pattern, not template)**

- 8/10 lead with a 1-paragraph definition of Claude Projects
- 7/10 include a numbered setup walkthrough
- 6/10 include screenshots
- 5/10 include a system-prompt example
- 0/10 include a Vault-style structured intake (white space)

**Content gaps no one covers well**

1. Multi-team/multi-brand Project organization (everyone shows the single-brand case)
2. How to handle Project context drift over time (the "stale Project" problem)
3. The exact system-prompt structure that marketers find useful
4. When NOT to use a Project (every article assumes Projects = better)

**Expertise signals we'd need**

- First-person screenshots of an actual Project setup
- A named example brand (real or fictional, but consistent and detailed)
- A linked Context Vault export (proof we use the system we're recommending)

**Brief**

- Working title: "How to Set Up Claude Projects for Marketing Teams (with the System Prompt We Actually Use)"
- Word count target: 1,800
- Must include: The 4 system-prompt sections we use; a 30-second screen recording; a downloadable system-prompt template; a "when NOT to use Projects" section nobody else has

**What NOT to write**

- "What is Claude Projects". Anthropic's docs already own this, don't compete
- A history of Anthropic / "AI for marketing" intro, wastes the first 400 words
- Generic prompt examples: the value is brand-specific structure, not generic prompts
- Word count padding ("there are many ways to..."): the brief is the value, not the length

How to use it well

1. Paste the actual top 10, not their URLs. Claude is dramatically better at synthesis from raw text than from links. Open each result, copy the body, paste it into the chat with the URL as a header.
2. Limit to one specific query. "How to use Claude" is too broad. "How to set up Claude Projects for a marketing team" is sharp enough to produce sharp output.
3. Trust the "what NOT to write" section. The pieces of content that rank are the ones with constraints, not the ones that try to cover everything. The discipline of refusing to write certain sections is half the battle.
4. Don't chase word count. Briefs that win specify unique value, not length. If the brief tells you 1,200 words, write 1,200, not 2,500.

Common questions

How is this different from a Surfer SEO or Frase brief?

Those tools optimize for keyword density and competitor matching. This brief optimizes for the gap. The difference shows up in what the brief tells you NOT to write: which is where most content tools are silent.

Does it work for AI Overview optimization specifically?

Yes, with one tweak: when the target query has an AI Overview surface, paste the AI Overview as one of the inputs alongside the 10 organic results. The brief will identify whether your content has a chance of being PULLED INTO the AI Overview vs being relegated to organic.

How long does it take?

15-25 minutes including the SERP capture. Most of the time is in pasting the 10 results, not in Claude's processing.

Can I run it for a query my brand has no business ranking for?

You can. The brief will work, but the output will tell you in the "expertise signals" section that you don't have the credibility to match. That's useful negative data, sometimes the right answer is "don't write this article."