These 25 Claude prompts for SEO cover the full workflow I run on this site: keyword research, content briefs, drafting and on-page optimization, technical audits, and link building. Claude won’t replace your rank tracker or crawler — it has no live index of search results — but paired with data you paste in, it compresses hours of SEO grunt work into minutes. Every template below is copy-paste ready.
TL;DR:
• 25 templates in 5 groups: keyword research (1–5), content briefs (6–10), on-page (11–16), technical SEO (17–21), links and digital PR (22–25).
• Claude has no live SERP access in a standard chat — paste exports from your SEO tools and let it do the analysis and writing.
• The biggest wins are in the middle of the workflow: briefs, rewrites, schema, and internal-link planning.
• Never publish AI-drafted claims or statistics without verifying them; that’s an E-E-A-T problem waiting to happen.
How to Use Claude Prompts for SEO Work
The mistake I made in my first month: asking Claude for keyword volumes and rankings. It doesn’t have them — numbers it produces from memory are guesses, and using them is how SEO work goes wrong. The workflow that actually holds up: export data from Ahrefs, Semrush, or Search Console, paste it in, and use Claude for the parts humans are slow at — clustering, brief writing, rewriting, and structured-data generation. With current models handling a 1M-token context window, an entire keyword export or a full page’s HTML fits comfortably in one message. If prompt structure is new to you, start with the ultimate Claude prompts library and Anthropic’s prompt engineering guide.
Keyword Research Prompts (1–5)
1. Keyword clustering: Here's a keyword export with volume and difficulty: [PASTE CSV ROWS]. Cluster by search intent, name each cluster, assign one primary keyword per cluster, and mark which clusters one page can serve vs. which need separate pages. Output as a table.
2. Intent classification: Classify each keyword as informational, commercial, transactional, or navigational, and state what page type best serves it (guide, comparison, product page, tool): [PASTE KEYWORDS]. Flag ambiguous ones where the SERP could go either way.
3. Long-tail expansion: Seed keyword: [KEYWORD]. Generate 30 long-tail variations a real [AUDIENCE] would type, grouped by: questions, comparisons, "best/top" modifiers, problem phrasings, and use-case phrasings. Mark the 5 most likely to convert and say why.
4. Topical gap analysis: Here are the topics we've covered: [PASTE URL/TITLE LIST]. Here are competitor topics: [PASTE LIST]. Identify gaps in our coverage, rank them by strategic fit with what we sell ([PRODUCT]), and flag competitor topics NOT worth copying.
5. SERP intent read: I'm pasting the top 10 titles and descriptions ranking for [KEYWORD]: [PASTE SERP]. What intent is Google rewarding? What format dominates (listicle, guide, tool, comparison)? What angle is missing that we could own?Content Brief and Outline Prompts (6–10)
6. Full content brief: Create a content brief for the keyword [KEYWORD]. Include: search intent, working title (≤60 chars), H2/H3 outline, questions to answer (from this PAA list: [PASTE]), semantic keywords to include, suggested word count with reasoning, and internal links from this list: [PASTE YOUR URLS].
7. Outline from the SERP: Here are the H2s from the top 3 ranking pages for [KEYWORD]: [PASTE HEADINGS]. Build an outline that covers the consensus subtopics, then add 2 sections none of them have, based on this angle: [YOUR UNIQUE ANGLE/DATA].
8. Title and meta batch: For a page targeting [KEYWORD], write 8 title tags (≤60 chars, keyword included, no clickbait) and 5 meta descriptions (≤155 chars, keyword + concrete benefit). Show character counts. Then pick your winner and defend it in one sentence.
9. FAQ section builder: For an article about [KEYWORD], write 6 FAQ questions real searchers ask (based on these PAA results: [PASTE]) with 2-4 sentence answers. Answers must be self-contained and quotable — no "as mentioned above".
10. Featured snippet target: The current snippet for [KEYWORD] is: [PASTE SNIPPET]. Write a better version: 40-55 words for a paragraph snippet, or a numbered list if the query implies steps. Direct answer in the first sentence, no preamble.On-Page Optimization Prompts (11–16)
11. Content refresh: Here's a page that's slipped from position [X] to [Y] for [KEYWORD]: [PASTE CONTENT]. Diagnose likely content issues (thin sections, outdated facts, missing subtopics vs. this current SERP: [PASTE]), then rewrite the 3 weakest sections.
12. Keyword integration pass: Here's my draft: [PASTE]. Primary keyword: [KEYWORD], secondary: [LIST]. Integrate them naturally — first 100 words, at least one H2, and body — WITHOUT changing my voice or stuffing. Show me a diff-style list of every change you made.
13. Internal link planner: Here's my article: [PASTE]. Here's my site's URL list with topics: [PASTE]. Suggest 4-6 internal links: exact anchor text, target URL, and the sentence where each fits. Descriptive anchors only — no "click here", no raw URLs.
14. E-E-A-T strengthener: Review this draft for E-E-A-T signals: [PASTE]. Mark every claim that needs a source, every place first-hand experience would strengthen it (suggest what kind), and any YMYL-adjacent statement that needs hedging or expert review.
15. Readability edit: Edit this for scannability without dumbing it down: [PASTE]. Target: paragraphs ≤4 lines, no sentence over 30 words, subheading every 150-250 words, active voice. Keep all facts and links intact. Return the full edited version.
16. Schema markup: Generate valid JSON-LD for this page: [PASTE PAGE SUMMARY + URL + AUTHOR + DATES]. Type: [Article/FAQPage/HowTo/Product]. Include only properties I've given you real values for — no invented ratings, reviews, or dates.Technical SEO Prompts (17–21)
17. Crawl triage: Here's an export from my site crawler: [PASTE ISSUES LIST]. Group by: critical (blocks indexing/ranking), important (dilutes equity), cosmetic. For each critical item, explain the fix in plain language a developer can act on, and estimate effort (S/M/L).
18. Redirect map: We're migrating [OLD SITE STRUCTURE] to [NEW STRUCTURE]. Here are the old URLs: [PASTE]. Build a 301 redirect map to the new URLs, flag chains, list URLs safe to let 404, and output the map as [.htaccess rules / nginx rules / CSV].
19. Robots and indexing review: Here's my robots.txt and a list of my noindexed URLs: [PASTE]. Identify anything that blocks pages we likely want indexed, anything crawlable we probably don't (filters, params, thin tags), and write the corrected robots.txt.
20. Core Web Vitals translation: Here's my PageSpeed Insights output: [PASTE]. Translate into a prioritized dev ticket list: issue, affected metric (LCP/CLS/INP), plain-English cause, suggested fix, expected impact (high/med/low). No generic advice — tie each item to my actual output.
21. Log-file sample analysis: Here's a sample of Googlebot hits from our server logs: [PASTE]. What's Google crawling most vs. what matters most (priority URLs: [LIST])? Identify crawl waste and 3 concrete crawl-budget improvements.Link Building and Digital PR Prompts (22–25)
22. Linkable asset ideas: We're in [NICHE] and can produce [DATA/RESOURCES WE HAVE]. Propose 5 linkable asset concepts (original data study, free tool, definitive resource), each with: the specific audience who'd link, why they'd link, and the outreach hook in one sentence.
23. Outreach email: Write a link outreach email to [SITE/AUTHOR] who wrote [THEIR ARTICLE]. Our asset: [YOUR PAGE + WHY IT ADDS VALUE TO THEIR PIECE]. Under 90 words, specific reference to their content in line one, no flattery padding, one clear ask. 2 variants.
24. Broken link candidates: Here's a list of dead pages in our niche and sites linking to them: [PASTE FROM YOUR TOOL]. For each, say which of our pages ([LIST]) is the closest replacement, or say "no fit" — don't force matches. Draft the one-line pitch per real fit.
25. Digital PR angles: From this dataset we own: [DESCRIBE/PASTE DATA], extract 5 story angles journalists in [BEAT] would cover. For each: headline a journalist would write, the single stat that carries it, and which publications to pitch.What Claude Can and Can’t Do for SEO
| Task | Claude alone | Claude + your tool data |
|---|---|---|
| Keyword volumes and difficulty | No — guesses only | Yes — paste exports for clustering and prioritization |
| Live rankings / SERP data | No live index in standard chat | Yes — paste SERPs for intent and gap analysis |
| Briefs, outlines, titles, metas | Good | Excellent with PAA + SERP context |
| Rewrites, readability, keyword integration | Excellent | Excellent |
| Schema, redirect maps, robots.txt | Good — verify before deploying | Best with your real URL lists |
For high-volume programmatic work — generating hundreds of metas or briefs — chat stops scaling and the Claude API takes over; the same prompts work with a script wrapped around them. And if you run the same SEO instructions in every session, store them once as custom instructions — my Claude system prompts setup guide shows how.
Common Mistakes With SEO Prompts
- Trusting Claude’s numbers. Search volumes, difficulty scores, and “current rankings” from the model’s memory are fabrications. Data comes from tools; analysis comes from Claude.
- Publishing unedited drafts at scale. Bulk unreviewed AI content is exactly what Google’s helpful content signals demote. Every page needs a human pass and something original.
- Deploying technical output blind. Test redirect rules and robots.txt in staging. One bad wildcard can deindex a section of your site.
- Keyword-stuffed rewrites. If you ask Claude to “add the keyword more”, it will comply past the point of sense. Give it a density ceiling and your voice samples.
- One-shot briefs with no SERP context. A brief written without pasting the actual SERP optimizes for an imagined query, not the real one.
Next Steps
Start with templates 1, 6, and 13 — clustering, briefs, and internal links are where the time savings show up first. When the content itself is the bottleneck, the Claude prompts for writers handle drafting and editing, and the marketing prompt library covers the promotion side. The complete collection lives on the prompts pillar page.
FAQ
Can Claude do keyword research on its own?
Not reliably. Claude has no live search index or keyword database, so any volume or difficulty figure it produces from memory is a guess. The working pattern is to export keywords from Ahrefs, Semrush, or Search Console and use Claude to cluster, classify intent, and prioritize.
Is Claude better than ChatGPT for SEO content?
In my testing Claude’s long-form drafts need less de-AI-ing — fewer stock phrases and better instruction-following on tone and structure. Both are capable; the deciding factor is usually which model matches your voice with less editing. Test both on one brief and compare.
Does Google penalize content written with Claude?
Google evaluates helpfulness, not authorship. AI-assisted pages with real expertise, accurate facts, and original insight can rank well; mass-produced unedited output is what gets demoted. Verify every claim and add first-hand experience before publishing.
Can Claude generate schema markup safely?
Yes — it produces clean JSON-LD for Article, FAQPage, HowTo, and Product types. Two rules: only include properties you gave it real values for, and validate the output with Google’s Rich Results Test before deploying.
How do I scale these prompts across hundreds of pages?
Move from chat to the Claude API. The same templates work programmatically — loop over your URL or keyword list, call the API with the template per row, and batch the output for human review. The Message Batches API cuts costs on large non-urgent jobs.