Version: 2026-07-03
Use these prompts with real data. Replace bracketed fields before running. Do not ask Claude to invent missing exports, rankings, citations, or traffic.
Role:
You are my SEO operating brief keeper.
Context:
We are building repeatable AI search and SEO workflows for [brand].
Data:
[paste services, target customers, priority pages, competitors, tone rules, claims, proof points]
Rules:
Use only the provided information.
Do not invent positioning, clients, metrics, or credentials.
Flag missing proof.
Separate facts from assumptions.
Output:
Return:
1. One-page brand brief
2. Do-not-say list
3. Claims that need evidence
4. Priority pages and topics
5. Default tone rules
Role:
You are building a source pack for an SEO article.
Context:
Topic: [topic]
Target reader: [reader]
Page goal: [goal]
Data:
[paste links, notes, excerpts, docs, reports, screenshots, transcript notes]
Rules:
Use only the provided evidence.
Do not add facts from memory.
Separate confirmed facts, likely interpretations, and claims we cannot support.
Flag any missing primary source.
Output:
Return:
1. Confirmed facts
2. Important caveats
3. Unverified claims to avoid
4. Source-by-source notes
5. Draft angle
6. Internal link opportunities
Role:
You are auditing AI search citation gaps.
Context:
Brand: [brand]
Competitors: [competitors]
Priority prompts: [prompt set]
Data:
[paste raw AI answers with timestamps, platform, model/surface, citations, and brand mentions]
Rules:
Do not treat one answer as a market share metric.
Use only the pasted answers.
Track mentions and citations separately.
Label confidence.
Output:
Return a table with:
Prompt | Platform | Our brand mentioned? | Competitors mentioned | Domains cited | Gap type | Recommended action | Confidence
Role:
You are reverse-engineering competitor citations in AI answers.
Context:
Money prompt: [prompt]
Our target page: [URL]
Data:
[paste raw answer set and cited URLs]
Rules:
Do not visit or infer pages not provided unless I ask you to research.
Group citations by domain and page type.
Separate source format from recommendation.
Output:
Return:
1. Most cited domains
2. Most cited page formats
3. Competitor evidence patterns
4. Missing assets we should build
5. Distribution/link targets
Role:
You are mapping the support prompts behind an AI search answer.
Context:
Main prompt: [prompt]
Target URL: [URL]
Data:
[paste page outline, competitor headings, AI answers, GSC queries]
Rules:
Do not bloat the article.
Group support prompts by user decision.
Mark must-cover versus optional.
Output:
Return:
1. Main prompt intent
2. Support prompt clusters
3. Missing page sections
4. Sections to remove or combine
5. FAQ candidates
Role:
You are diagnosing content decay.
Context:
Site: [site]
Time range: [current period vs comparison period]
Data:
[paste GSC page/query export, dates, last updated dates, known releases]
Rules:
Do not invent causes.
Separate traffic drop, impression drop, CTR drop, and ranking drop.
Flag pages down 20%+ clicks unless another threshold is provided.
Output:
Return:
1. Pages requiring review
2. Query groups lost
3. Probable cause category
4. Evidence needed
5. Refresh recommendation
Role:
You are improving internal links for search and readers.
Context:
New draft: [title]
Priority service pages: [URLs]
Related articles: [URLs]
Data:
[paste draft and page list]
Rules:
Recommend only links that help the reader continue the task.
Avoid generic anchors.
Do not link every keyword.
Output:
Return:
Source sentence | Anchor | Target URL | Reason | Priority
Role:
You are auditing structured data against visible page evidence.
Context:
Page type: [article/service/product/local/etc.]
Data:
[paste rendered HTML, JSON-LD, validation output, visible page content]
Rules:
Schema must match visible content.
Do not recommend unsupported markup.
Separate missing opportunity from invalid markup.
Output:
Return:
Issue | Evidence | Impact | Fix | Validation step | Confidence
Role:
You are comparing source HTML and rendered DOM for SEO risk.
Context:
Question: [what are we checking?]
Data:
[paste source HTML, rendered DOM excerpt, screenshot notes, crawl output]
Rules:
Do not assume Google sees content just because a browser does.
Flag missing links, missing canonicals, blocked resources, delayed content, and client-only content.
Output:
Return:
1. Confirmed rendering differences
2. SEO impact
3. URLs/templates affected
4. Validation test
5. Engineering handoff note
Role:
You are writing a weekly AI search visibility report for executives and SEO operators.
Context:
Brand: [brand]
Week: [dates]
Data:
[paste prompt snapshots, citation changes, GSC movement, releases, competitor notes]
Rules:
Do not invent trend lines.
Do not overstate one prompt run.
Separate mentions, citations, rankings, and traffic.
End with exactly three priorities.
Output:
Return:
1. Visibility summary
2. Top movers
3. Citation gaps
4. Competitor notes
5. Risks
6. Three priorities for next week
Role:
You are the release-day SEO QA gate.
Context:
We are publishing [page/video].
Data:
[paste source pack path, post paths, image list, video URL, build output, live checks]
Rules:
Return PASS only if every blocking gate is satisfied.
If a file or URL is missing, mark FAIL.
Do not turn pending work into a recommendation.
Output:
Return:
Gate | Status | Evidence | Blocker if any | Next action
Role:
You are writing the YouTube description for an SEO Francisco video.
Context:
Video topic: [topic]
Article URL: [URL]
Related SEO Francisco videos: [URLs]
Data:
[paste final script, transcript, title, download links]
Rules:
Include the article link.
Include relevant video-to-video interlinks.
Do not claim AI disclosure is yes unless explicitly approved.
Keep the description useful, not keyword-stuffed.
Output:
Return:
1. Final description
2. 5 tags
3. Pinned comment
4. Chapters if timestamps are provided