In short: A prompt library is a set of reusable instructions for AI assistants, organized by task. The 30 prompts below cover the full SEO and GEO workflow: keyword research, content planning, writing, on-page optimization, structured data, and checking how the AIs talk about your brand. Copy them, replace the variables in [brackets], and give the assistant as much context as you can.
Most people get average answers from ChatGPT or Claude because they ask average questions. A good prompt library fixes that: each prompt states a role, the context, the task and the expected format. Save the ones you use in a shared document so your whole team gets the same quality.
How do you write a good SEO prompt?
Give a role, the business context, a precise task and the output format. The assistant knows SEO in general; it knows nothing about your site, your market or your competitors unless you tell it.
- Role: "You are a senior SEO strategist."
- Context: your site, offer, audience, country and language.
- Task: one action per prompt, with constraints.
- Format: a table, a list, an outline, a word count.
Our glossary entry on prompt engineering goes deeper into these techniques.
Keyword research prompts
- Seed list. "You are an SEO strategist. My site sells [product] to [audience] in [country]. List 40 search queries my buyers type in Google, grouped by intent (informational, commercial, transactional), in a table."
- Long tail. "For the keyword [keyword], give 25 long-tail variants that a beginner would type, with the likely intent of each."
- Questions. "List the 20 most common questions people ask about [topic], phrased exactly as they would type them."
- Intent check. "Here are my keywords: [list]. For each one, say whether the searcher wants a guide, a comparison, a product page or a local result, and which page type I should create."
- Search Console quick wins. "Here is a Search Console export: [paste]. Find the queries with many impressions, a position between 5 and 20 and a low click-through rate, and suggest an action for each."
An assistant estimates; it does not measure. Check volumes and difficulty in a keyword tool before you commit, or let Sorank's Keyword Research attach live volume, difficulty and intent to each keyword.
Content planning prompts
- Topic clusters. "Group these keywords into topical clusters, with one pillar page and its supporting articles for each cluster: [list]."
- Cannibalization. "Here are my existing page titles: [list]. Here are new keyword ideas: [list]. Flag every new keyword that an existing page already targets."
- Editorial calendar. "Turn these clusters into a 3-month calendar of 3 articles per week, starting with the easiest topics and alternating clusters."
- Competitor gaps. "Here are the article titles of a competitor: [list]. Which topics do they cover that I do not, ranked by business value for [my offer]?"
Writing prompts
- Content brief. "Write a content brief for [keyword]: search intent, target reader, H2 and H3 outline phrased as questions, points each section must answer, data to include, and internal links to [pages]."
- Answer-first intro. "Write a 60-word introduction for an article on [topic] that answers the main question in the first sentence."
- Section draft. "Write the section [H2] in 200 words. Start with a one-sentence direct answer, then explain with a concrete example. Tone: [tone]. Address the reader as [you, formal or informal]."
- FAQ. "Write 5 FAQ questions and 50-word answers for an article on [topic], each answer complete on its own."
- Brand voice. "Here are three paragraphs from our site: [paste]. Describe our tone, vocabulary and sentence style in 8 rules, then rewrite this paragraph following them: [paste]."
- Fact check. "List every factual claim in this text, and for each one say whether it needs a source and what kind of source would support it: [paste]."
Need consistent tone across hundreds of articles? Brand Voice in Sorank applies your tone, terminology and call to action to every article automatically.
On-page optimization prompts
- Title tags. "Write 5 title tags under 60 characters for a page targeting [keyword], with the keyword at the start."
- Meta descriptions. "Write 3 meta descriptions under 155 characters for [page], each with a clear benefit and the keyword."
- Content refresh. "Here is an article that lost traffic: [paste]. Compare it with what a reader searching [keyword] in 2026 expects, and list what to add, update or remove."
- Internal links. "Here is my article: [paste] and my list of pages with URLs: [list]. Suggest 5 internal links with natural anchor text and where to place them."
- Image alt text. "Write descriptive alt text under 120 characters for these images: [descriptions]."
Structured data prompts
- FAQ schema. "Turn these questions and answers into valid FAQPage JSON-LD: [paste]."
- Article schema. "Write Article JSON-LD for this page with headline, author, dates, image and publisher: [details]."
- Organization schema. "Write Organization JSON-LD for [company] with logo, URL and sameAs links to [profiles]."
Always validate the output with Google's Rich Results Test before publishing: assistants sometimes invent properties.
GEO and AI visibility prompts
- Buyer questions. "List 25 questions a buyer of [product] in [country] would ask an AI assistant before choosing a provider, labeled understanding, benefits, comparison or quality."
- Brand check. "Which [category] providers would you recommend for [use case] in [country]? Explain why for each one." Run it in several AIs and note whether your brand appears.
- Perception. "What do you know about [brand]? What are its strengths, weaknesses and alternatives?"
- Source discovery. "For the question [question], which websites and sources would you rely on to answer?"
- Citability rewrite. "Rewrite this section so each paragraph answers one question completely, starts with the answer, and includes one sourced figure: [paste]."
- Comparison table. "Build a neutral comparison table of [options] on price, use case, strengths and limits, for an article on [topic]."
- Entity consistency. "Here is how we describe our company on our site, LinkedIn and directories: [paste]. List the inconsistencies an AI could pick up."
The research behind GEO found that adding sources, quotations and statistics lifts visibility in generative answers by up to about 40% (Princeton GEO study), which is why prompts 28 and 29 matter.
What prompts alone cannot do
An assistant answers once, from what it knows; SEO and GEO need data, repetition and authority. Prompts 24 to 27 give you a snapshot of a single conversation. Answers change by model, by week and by phrasing, so a reliable picture needs the same questions asked every week to every AI, with mentions counted.
That is what Sorank automates: AI Prompt Research writes and sorts the questions of your market, AI Mention Tracking measures your brand on each AI every week, and articles and backlinks follow on autopilot. Prefer to work in a conversation? SEO Chat answers SEO questions using your own site and data.
Conclusion
A good prompt library saves time and raises the floor of every answer, provided you give context and verify the facts. Start with the five keyword prompts and the GEO prompts, save your best versions, and share them with your team.
When you are ready to go from prompts to results, start a 3-day free trial of Sorank.