ChatGPT (and Claude) have become the default assistants for SEO professionals, handling tasks from keyword clustering to competitor analysis to content drafting. Yet the quality of output depends entirely on the quality of the prompt. A vague request to “write a blog post” yields generic, unusable content; a precisely crafted prompt with context and specifications yields material you can actually use. This guide provides ten prompts SEO professionals use daily. Each is field-tested and works with ChatGPT, Claude, or similar language models. Treat them as starting points, not finished products — you still need to fact-check output, adjust for brand voice, and verify that generated content aligns with your actual SEO strategy and Google’s policies on scaled content.
Prompt 1: keyword research and semantic clustering
Use when you have a seed keyword and need to identify long-tail variations grouped by search intent. Saves hours of manual research and helps plan comprehensive content strategies.
Prompt: “I am targeting the seed keyword ‘[YOUR KEYWORD HERE]’. Generate a table with 25 long-tail keyword variations related to this term. For each keyword, include: the keyword itself, estimated search intent (informational, transactional, or commercial), a potential blog post title that would rank for this keyword, and an estimated relative volume tier (high, medium, low) based on typical patterns for this topic. Format as a markdown table. The keywords should range from very specific long-tails to broader variations.”
Important: don’t trust LLM-generated absolute search volumes — they are pattern-matched estimates, not real data. Always cross-reference with Google Keyword Planner, Ahrefs, or Semrush before bidding decisions or content prioritization. The LLM’s value is the variation generation and intent classification, not the volume numbers.
Prompt 2: content brief generation
When you need a detailed brief without spending hours researching competitor content, this prompt generates a structured outline that guides writers without being overly prescriptive.
Prompt: “Create a comprehensive content brief for the keyword ‘[TARGET KEYWORD]’. Include: 1) Keyword Overview (search intent, audience type, typical pain points), 2) Content Structure (suggested H1, H2 headings, content flow), 3) Key Topics to Cover (15-20 subtopics top-ranking pages typically address), 4) FAQ Section (10 common questions the audience asks), 5) Recommended Word Count and Reading Time, 6) Target Audience and Tone. Make the brief actionable and specific to this keyword. Assume the writer has never researched this topic; the brief should contain everything they need.”
For higher accuracy, paste the URLs or content of the top 3 ranking pages alongside the prompt. The LLM then bases the brief on actual current SERP content rather than its training data, which may be outdated.
Prompt 3: meta description writing
Meta descriptions don’t directly affect rankings but influence CTR from search results. This prompt generates multiple variations to test.
Prompt: “Generate 5 unique meta descriptions for the following content: Title: ‘[YOUR PAGE TITLE]’, Keyword: ‘[TARGET KEYWORD]’, Page Summary: ‘[ONE-SENTENCE SUMMARY]’. Each should be 150-160 characters (including spaces), include the target keyword naturally, and include a compelling call-to-action or benefit. Prioritize click-through appeal over keyword stuffing. Number each variation and show the character count.”
Test variations in Search Console after deploying. Don’t trust the LLM’s character count claim blindly — verify each variation manually because LLMs sometimes miscount, especially around emoji or special characters that count as multiple bytes in some search interfaces.
Prompt 4: FAQ schema generation
FAQPage schema markup makes content eligible for FAQ rich results in some queries. This prompt generates ready-to-implement structured data in JSON-LD.
Prompt: “Generate 10 FAQ questions and answers for the topic ‘[YOUR TOPIC]’. Format the output as JSON-LD schema markup ready for WordPress or any website. Each answer should be 50-150 words, clear and concise. Questions should reflect actual search queries users perform for this topic. After the JSON-LD code, provide a markdown version of the same FAQs for use in the blog post itself. Ensure both versions contain the exact same content.”
Critical: FAQPage schema content must actually appear visibly on the page. Google’s documentation explicitly states that hidden FAQ content marked up with schema is a violation. The markdown version of the same FAQs needs to be visible to users, not just baked into JSON-LD. Validate the resulting JSON with the Rich Results Test before deploying.
Prompt 5: topic clustering and content mapping
When you have dozens of keywords, clustering helps you understand which should be covered by single comprehensive pillar pages and which should be standalone content.
Prompt: “I have the following 40 keywords related to ‘[TOPIC AREA]’: [PASTE YOUR KEYWORD LIST]. Group these keywords into 5-7 topic clusters where each cluster represents a pillar page topic. For each cluster, designate one keyword as the pillar page target and list the remaining keywords as supporting cluster content. For each cluster, suggest a pillar page title and describe the relationship between pillar and cluster content. Provide a recommended internal linking strategy to connect cluster content to the pillar page.”
The output is a content roadmap. Validate by checking whether the clusters reflect actual search intent groups (some keywords ChatGPT clusters together may have very different SERPs in reality). Spot-check a few keywords against actual Google results before committing the strategy.
Prompt 6: competitor content gap analysis
Use when you want to identify topics competitors cover that you don’t.
Prompt: “Here are the top 3 ranking URLs for ‘[TARGET KEYWORD]’: [URL 1], [URL 2], [URL 3]. Here is my current content on this topic: [PASTE YOUR EXISTING CONTENT OR URLs]. Identify: 1) Topics or angles competitors cover that I don’t, 2) Specific subtopics competitors treat in depth where I’m shallow, 3) Questions competitors answer that mine doesn’t, 4) Format differences (e.g., they use comparison tables, I don’t). Output as a prioritized gap list with specific recommendations for each gap.”
Pair this with a manual SERP review — the LLM may not have actually fetched the URLs (depending on which model and tools are enabled). For ChatGPT, ensure browsing is enabled or paste the relevant content directly. The output is only as accurate as the input.
Prompt 7: outreach email for link building
Use when reaching out to a publication, blog, or business that linked to a competitor’s page or covered a similar topic.
Prompt: “Write a personalized outreach email for link building. Recipient: [PERSON NAME] at [PUBLICATION]. They recently published this article: [URL OR ARTICLE TOPIC]. I am offering: [YOUR RESOURCE: a guide, original research, tool, etc.] which would add value to their article because [SPECIFIC REASON]. The tone should be professional but warm, under 150 words, and end with a clear call to action. Avoid generic phrases like ‘I love your work’ or ‘just wanted to reach out’. Be specific about what I bring.”
Personalize the LLM output further before sending — recipients can tell when an email was AI-templated. Edit the LLM draft to add a specific detail you noticed in their work that the LLM couldn’t have known.
Prompt 8: technical SEO audit interpretation
When you have a Screaming Frog or Ahrefs export with hundreds of issues, this prompt prioritizes them.
Prompt: “I have the following technical SEO audit findings from [TOOL NAME]: [PASTE TOP-LEVEL SUMMARY OR KEY ISSUES]. Categorize these into: 1) Critical (blocks crawling or indexing — fix this week), 2) High (impacts rankings — fix this month), 3) Medium (best practice — fix this quarter), 4) Low (nice to have — fix when convenient). For each Critical and High issue, provide a one-line explanation of the business impact and a one-line action item.”
The LLM is reasonably good at categorization but not always at business impact estimation. Sanity-check the priority order against your specific business context — a “Critical” 404 on a page that hasn’t received traffic in three years matters less than a “High” Quality Score issue on your top-converting page.
Prompt 9: content refresh recommendations
Use on existing content that has lost rankings or stopped converting.
Prompt: “Here is my existing article: [PASTE FULL ARTICLE OR URL IF BROWSING IS ENABLED]. It used to rank well for ‘[TARGET KEYWORD]’ but has lost positions. Top ranking competitors are: [URLs]. Identify: 1) Outdated information or examples that should be updated to 2026 standards, 2) Missing topics competitors now cover, 3) Structural improvements (better headings, FAQ section, comparison tables), 4) Specific sentences or paragraphs to rewrite for clarity or freshness, 5) Internal and external links to add. Format as a prioritized refresh checklist.”
Refresh recommendations from LLMs are useful for direction but not for facts. Verify any specific claim or statistic the LLM suggests adding — hallucinated stats inserted into refreshed content are how previously-good articles become factually wrong articles.
Prompt 10: title tag A/B testing variations
Use when you want to test multiple title variations on important pages.
Prompt: “Generate 8 title tag variations for: Page topic ‘[TOPIC]’, target keyword ‘[KEYWORD]’, target audience ‘[AUDIENCE]’. Each title should be under 60 characters and include the target keyword naturally. Vary the angle: 1 number-based (‘Top 10’, ‘5 Steps’), 2 question-based, 1 with negative framing (‘Mistakes to Avoid’), 1 with year (‘2026 Guide’), 1 benefit-focused (‘How to Save’), 1 with curiosity gap, 1 with authority signal (‘Expert Guide’). Show character count for each. Note which is best suited for which audience type.”
Test variations in Search Console (or via dedicated A/B testing tools like SEOTesting or Otto) and measure CTR before deciding the winner. Don’t accept the LLM’s “best for which audience” claim without measurement — A/B testing real users beats LLM intuition.
How to use these prompts effectively
Treat the LLM as a junior assistant, not an oracle. The output requires editing and verification on every prompt. The value of the LLM is generating volume of variations and structured first drafts; the value of you is selecting, fact-checking, and adapting to your specific brand and audience.
Always provide context. The more specific information you give the LLM (target audience, brand voice, current content, competitor URLs), the better the output. Vague prompts produce vague output.
Verify factual claims. LLMs hallucinate. Any specific number, statistic, claim, or attribution generated by the LLM needs verification before publication. Google’s March 2024 spam policies explicitly target scaled content abuse including unsupervised AI content; published LLM output without editorial review is increasingly risky.
Use these prompts as templates. Copy them, customize the brackets with your specifics, save the customized versions for reuse, and iterate. Over time, you’ll develop your own variations that work better for your specific niche.
Building a prompt library
Beyond individual prompts, building a prompt library that your team can use consistently multiplies productivity. Store the customized versions of these prompts in a shared document, with notes on when to use each, what context to include, and what to verify in the output. New team members can use the library to produce consistent SEO outputs without each person reinventing the prompts.
Keep the library in version control if you have engineering capacity. Prompts evolve as LLM capabilities change — what worked on GPT-3.5 may produce different output on GPT-5 or Claude Opus 4.7. Versioning lets you track which prompt version produced which output, important for reproducibility.
Conclusion
Prompts are the new keyword research — the more carefully crafted, the more useful the output. The ten prompts above cover the most common SEO use cases: keyword research, content briefs, meta descriptions, FAQ schema, topic clustering, gap analysis, outreach, technical audit interpretation, content refresh, and title tag testing. Customize them with your specific context, verify the output, and iterate. The accounts that win at this level treat LLMs as productivity multipliers under human supervision, not as autopilots.
LaFactory builds prompt libraries matched to client SEO workflows, with verification checklists baked in to keep AI productivity safe under Google’s scaled-content policies. Contact us to scope a prompt library and AI workflow setup for your team.