SEO Prompts for LLMs

Stojan

Updated on March 1, 2026

SEO Prompts for LLMs

Most SEO tasks follow repeatable patterns. Writing meta descriptions, structuring headers, brainstorming keywords, and building content briefs all start from a consistent template. That makes them a natural fit for LLMs.

This collection is part of a growing AI prompt library organized by task. The 18 prompts below cover six common SEO jobs, from meta descriptions to internal link planning.

Each prompt is ready to copy into ChatGPT, Claude, or Gemini. Fill in every bracketed placeholder with details about your page, topic, or brand for the best results.

You can use the prompt groups in sequence or pick individual ones. Meta description and title tag prompts handle on-page elements, while keyword research, content brief, header, and internal linking prompts cover planning and structure.

How to Use These Prompts

Every prompt includes bracketed placeholders like [KEYWORD] or [TARGET AUDIENCE]. Replace these with your own details before submitting. The more specific your inputs, the more useful the output.


Start with a single prompt and review the output before running the next one. LLMs produce better results when you give feedback on the first attempt and ask for a revision. This iterative approach matters more than finding the “perfect” prompt on the first try.

These prompts are model-agnostic. They work in ChatGPT, Claude, and Gemini without modification. Some models follow constraints like character limits more reliably than others, so test your preferred model and adjust if needed.

A few general guidelines apply to every prompt in this collection. First, specify your audience. “B2B SaaS marketers” produces tighter output than “marketers.”

Second, include hard constraints. Character limits, word counts, and format requirements keep the output usable without heavy editing. Third, add context by pasting in existing content, competitor titles, or brand guidelines when the prompt allows it.

If you are new to structuring prompts for better output, the fundamentals of structuring prompts for better output apply to every example here. Small changes to how you frame a request can change the quality of the result significantly.

Meta Description Prompts

Meta descriptions appear below your page title in search results. They do not directly affect rankings, but they strongly influence click-through rates. A well-written description can double or triple clicks compared to a generic one.

Writing them by hand for dozens or hundreds of pages is tedious. The format is constrained enough that an LLM can produce a solid first draft in seconds. You still need to review and edit, but the time savings compound quickly across a large site.

Google truncates meta descriptions at roughly 155 characters. Anything beyond that gets cut off. The prompts below enforce character limits so the output is ready to paste into your CMS without trimming.

Write a Meta Description from a Keyword

Use this when starting a new page and you only have a target keyword.

Prompt
Write a meta description for a web page targeting the keyword [KEYWORD]. The page is about [BRIEF TOPIC DESCRIPTION]. The description must be 135-155 characters, include the keyword naturally, and end with an implied benefit or call-to-action. Do not start with “Discover” or “Learn how to.”
Expected output
A single meta description ready to paste into your CMS

Rewrite an Underperforming Meta Description

Use this when Google Search Console shows a page with high impressions but low click-through rate. The prompt includes your current CTR to give the LLM context about the problem.

Prompt
Here is the current meta description for a page about [TOPIC]: “[CURRENT META DESCRIPTION]” The page ranks for [KEYWORD] but has a click-through rate of [CTR%]. Rewrite the meta description to be more compelling. Stay under 155 characters. Focus on a specific benefit or outcome the reader will get. Avoid generic language.
Expected output
A revised meta description with a stronger value proposition

Generate Multiple Meta Description Variants

Use this when you want to A/B test or pick the strongest option from several angles.

Prompt
Write 5 different meta descriptions for a blog post titled “[POST TITLE]” targeting the keyword [KEYWORD]. Each description must be 135-155 characters. Vary the angle: use a statistic in one, a question in one, a direct benefit in one, a comparison in one, and a how-to framing in one. Label each with its angle.
Expected output
Five labeled variants with distinct approaches, making it easy to pick a winner

Title Tag Prompts

Title tags carry significant ranking weight. They appear as the clickable blue headline in search results and in browser tabs. A strong title tag puts the primary keyword near the front and gives the reader a clear reason to click.

The challenge with title tags is balancing competing goals. You need the primary keyword for rankings, a modifier for differentiation, and enough specificity that users know what the page covers. Doing all three in under 60 characters takes more thought than most people expect.

Most SEO tools recommend keeping title tags between 50 and 60 characters. Going longer risks truncation in search results.

Going too short wastes an opportunity to include modifiers or differentiators. The prompts below enforce these limits by default.

Create a Title Tag from Scratch

Use this for new pages when you have a primary keyword and topic but have not settled on a title yet.

Prompt
Write an SEO title tag for a page about [TOPIC]. The primary keyword is [KEYWORD]. Requirements: 50-60 characters, keyword appears in the first half, reads naturally to a human, includes a modifier like “Guide,” “Examples,” or a number. Do not use clickbait or all caps.
Expected output
A single title tag with the keyword front-loaded

Optimize an Existing Title Tag

Use this when a page is ranking on page two and you want to test a title change. Including competitor titles gives the model something to differentiate against.

Prompt
The current title tag is: “[CURRENT TITLE]” The page targets [KEYWORD] and currently ranks position [POSITION] in Google. The top 3 competing titles are: 1. [COMPETITOR TITLE 1] 2. [COMPETITOR TITLE 2] 3. [COMPETITOR TITLE 3] Suggest 3 alternative title tags (50-60 characters each) that differentiate from competitors while keeping the primary keyword. Explain what makes each one different.
Expected output
Three alternatives with reasoning, so you can choose based on competitive positioning

Generate Title Tags for a Content Cluster

Use this when planning a group of related pages and you need consistent, non-overlapping titles. This prevents keyword cannibalization within your own site.

Prompt
I’m creating a content cluster around the topic [MAIN TOPIC]. The hub page targets [HUB KEYWORD]. The supporting pages target these keywords: 1. [KEYWORD 1] 2. [KEYWORD 2] 3. [KEYWORD 3] 4. [KEYWORD 4] Write a title tag for each page (50-60 characters). Make sure no two titles could be confused for the same content. The hub title should be broader; supporting titles should be more specific.
Expected output
A coordinated set of title tags with clear differentiation between pages

Keyword Research Prompts

LLMs cannot access live search volume data. They will not replace tools like Ahrefs or SEMrush for data-driven keyword research. Where LLMs excel is brainstorming keyword variations, grouping terms by search intent, and spotting content gaps you might overlook.

The gap between human and LLM brainstorming is significant. A person might generate 10 to 15 keyword ideas in 20 minutes. An LLM can produce 50 or more variations in seconds, covering question formats, comparison phrases, and niche modifiers.

Use these prompts at the ideation stage before validating with a keyword tool. The model generates candidate keywords based on its training data. You then check actual volume and difficulty in a dedicated tool.

Knowing how LLM tokenization works helps here. Large keyword lists consume token capacity fast, and the model needs enough room to process your input and generate a complete response.

Brainstorm Long-Tail Keywords

Use this to expand a seed keyword into specific phrases that reflect how real people search. Long-tail keywords often have lower competition and higher conversion rates than broad terms.

Prompt
I’m writing about [BROAD TOPIC] for [TARGET AUDIENCE]. My seed keyword is [SEED KEYWORD]. Generate 20 long-tail keyword variations (3-6 words each). Group them into categories: informational intent, commercial intent, and comparison intent. Focus on questions and specific use cases that a beginner would search for.
Expected output
A categorized list of 20 keyword ideas grouped by search intent

Map Keywords to Search Intent

Use this after you have a keyword list and need to decide what type of content each term requires. Matching content format to intent is one of the strongest ranking signals you can control.

Prompt
Here is a list of keywords related to [TOPIC]: [PASTE KEYWORD LIST] For each keyword, classify the search intent as informational, navigational, commercial, or transactional. Then recommend the best content format (blog post, comparison page, product page, or landing page). Present as a table with columns: Keyword, Intent, Recommended Format.
Expected output
A table mapping each keyword to intent and content format

Identify Content Gaps

Use this when you have existing content and want to find missing topics. Content gaps are pages your audience expects to find on your site but that do not exist yet.

Prompt
I run a website about [NICHE]. Here are the topics I’ve already covered: [LIST OF EXISTING ARTICLES OR TOPICS] Based on this coverage, identify 10 topic gaps that a visitor interested in [NICHE] would expect to find. For each gap, suggest a primary keyword and explain why the topic is missing. Prioritize topics that would connect logically to my existing content.
Expected output
Ten topic gap suggestions with keyword recommendations and rationale

Content Brief Prompts

A content brief tells a writer what to cover, how to structure the piece, and what separates it from competing pages. Good briefs save more time than any other part of the SEO content process. Without one, writers guess at scope and structure, which leads to rewrites.

Creating a brief manually takes 30 to 60 minutes per article. That includes reviewing competitors, deciding on structure, and planning internal links.

An LLM can produce a comparable brief in under five minutes. Output quality depends on how much context you provide. The briefs below ask for specific inputs like competitor summaries and existing site pages.

These prompts generate briefs ranging from simple outlines to detailed plans with competitive analysis. Pick the prompt that matches how much planning your article needs.

Generate a Full Content Brief

Use this when planning a new article from scratch and you want a complete starting plan.

Prompt
Create a content brief for an article targeting the keyword [KEYWORD]. The target audience is [AUDIENCE]. Include: Suggested title tag (50-60 characters) Meta description (135-155 characters) H2 and H3 outline with brief descriptions of what each section covers 3-5 key questions the article must answer Suggested word count range 2-3 competing articles to reference (describe the type, don’t generate URLs) Unique angle: what should this article do that competitors probably don’t
Expected output
A structured brief ready to hand to a writer or use as a writing plan

Create a Brief from Competitor Analysis

Use this when you have already reviewed competing articles and want to outline something stronger. The parity vs. differentiation labels help you see where you are matching the market and where you are pulling ahead.

Prompt
I’m writing an article targeting [KEYWORD]. Here’s what the top-ranking competitors cover: Competitor 1: [SUMMARY OF WHAT THEY COVER] Competitor 2: [SUMMARY OF WHAT THEY COVER] Competitor 3: [SUMMARY OF WHAT THEY COVER] Create a content brief that covers everything the competitors address AND adds value they miss. Suggest an H2/H3 outline. For each section, note whether it matches competitors (parity) or goes beyond them (differentiation). Mark differentiation sections clearly.
Expected output
An outline with clear parity and differentiation labels per section

Build a Content Brief with Internal Linking Plan

Use this when your article needs to connect to an existing network of pages on your site. Building links into the brief means they get written naturally into the first draft instead of bolted on after.

Prompt
I’m writing an article targeting [KEYWORD] for [WEBSITE]. Here are related pages on the site: [LIST OF RELATED PAGES WITH THEIR URLs AND TOPICS] Create a content brief with an H2/H3 outline. For each section, suggest where to place internal links to the related pages listed above. Use natural anchor text, not “click here.” Each link should appear once. The first internal link should appear within the first 300 words.
Expected output
A content brief with a built-in internal linking plan

Header Structure Prompts

Header hierarchy tells search engines and readers how your content is organized. A clear H2/H3 structure makes articles easier to scan and gives Google multiple entry points for ranking individual sections.

Google can pull any well-structured section into featured snippets or passage-based rankings. That means a single article can rank for its main keyword and for several long-tail queries addressed by individual sections. The header structure determines which sections qualify.

Poor header structure is one of the most common problems in SEO content. Vague headers like “Overview” or “More Information” waste ranking opportunities. Descriptive headers that preview the section content perform better in both rankings and user engagement.

Create a Header Outline from a Keyword

Use this at the start of a new article when you need an organized structure to write against.

Prompt
Create an H2 and H3 header outline for a [WORD COUNT]-word article targeting [KEYWORD]. The audience is [TARGET AUDIENCE]. Requirements: 5-7 H2 sections 2-3 H3s under the most complex H2s Include the primary keyword or a close variant in 2-3 H2s Headers should preview the section content (no vague headers like “Overview” or “Details”) Add a brief 1-sentence description under each header explaining what that section covers
Expected output
A complete outline with descriptive headers and section summaries

Restructure a Messy Draft

Use this when you have a completed draft with poor organization. The model identifies redundant sections and suggests a logical flow.

Prompt
Here is a draft article about [TOPIC]: [PASTE DRAFT TEXT] The target keyword is [KEYWORD]. Reorganize the content into a clear H2/H3 structure. Group related points under logical headers. Remove any redundant sections. Suggest where to add new content if any major subtopics are missing. Present the restructured outline with brief notes on what goes where.
Expected output
A reorganized header structure with placement notes for existing content

Featured snippet positions in Google pull content from well-structured pages. If you can see the current snippet for your target query, you can format your content to match and compete for position zero.

Prompt
I want to rank for the featured snippet for [QUERY]. The current featured snippet is a [PARAGRAPH/LIST/TABLE] that says: “[CURRENT SNIPPET CONTENT]” Create an H2/H3 structure for a section of my article that directly answers this query. Place the most direct answer in the first sentence after the H2. Follow it with supporting detail. Format the answer as a [PARAGRAPH/NUMBERED LIST/TABLE] to match the current snippet format.
Expected output
A header structure and answer format designed to win a specific featured snippet

Internal Linking Prompts

Internal links help search engines understand your site architecture and pass ranking authority between pages. They also keep readers on your site longer by connecting related content. Despite their importance, internal linking is one of the most neglected parts of on-page SEO.

Most sites have hundreds of missed internal linking opportunities. Pages that should link to each other simply do not, because no one audited them after publishing.

LLMs are good at spotting these connections. They can process an entire article alongside a list of related pages, then suggest natural link placements.

These prompts help you audit existing pages for missed links, generate varied anchor text, and plan link placement before writing. The model you choose affects output quality, and each handles SEO tasks differently.

Find Internal Linking Opportunities

Use this to audit an existing article and spot places where links to other pages on your site would fit naturally.

Prompt
Here is an article from my website: [PASTE ARTICLE TEXT] Here are other pages on my site with their topics: [LIST OF PAGES: TITLE – URL – TOPIC] Identify places in the article where I should add internal links to these pages. For each suggestion, provide: the sentence where the link should go, the anchor text to use, and the destination URL. Only suggest links that feel natural in context. Each URL should be linked once.
Expected output
A list of specific link placement suggestions with anchor text and destinations

Generate Anchor Text Variations

Repeating the same anchor text across your site looks unnatural to search engines. Google’s guidelines recommend varied, descriptive anchor text. Use this prompt when you link to the same page from multiple articles and need fresh phrasing each time.

Prompt
I need to link to a page about [TOPIC] with the URL [URL] from 5 different articles. The page’s primary keyword is [KEYWORD]. Generate 5 different anchor text options. Include: 1 exact-match keyword anchor 2 natural variations of the keyword 2 descriptive/contextual anchors that don’t use the keyword at all Each anchor should be 2-5 words.
Expected output
Five diverse anchor text options for the same destination page

Plan Internal Links for a New Article

Use this before you start writing so links are part of the content from the start, not an afterthought added during editing.

Prompt
I’m planning an article about [TOPIC] targeting [KEYWORD]. Here are the existing pages on my site: [LIST OF PAGES: TITLE – URL – BRIEF DESCRIPTION] Suggest 8-12 internal links I should include in this article. For each one, specify: which section of the article it belongs in, the recommended anchor text, and the destination URL. Place the first link within the first 300 words. Include at least one link to a page in a different content category.
Expected output
A pre-writing link plan organized by article section

Tips for Better Results

The prompts above work with any major LLM, but small adjustments to how you use them make a noticeable difference in output quality.

Give the Model Specific Context

Include real data whenever possible. Paste your actual Search Console data, competitor titles, or existing content into the prompt. A prompt with your actual page title, keyword, and CTR will always outperform a vague request.

Tell the model your brand voice. LLMs do not know your tone unless you specify it.

Add something like: “Our tone is direct and practical. We write for small business owners, not enterprise marketers.”

Character and word count limits matter for title tags and meta descriptions. Always specify exact limits in the prompt. Without constraints, LLMs tend to run long.

Google truncates titles past 60 characters and descriptions past 155.


LLMs do not have access to live search data. They cannot tell you actual search volume, keyword difficulty, or current rankings. Use the keyword and content brief prompts for brainstorming, then validate every suggestion with a dedicated SEO tool like Google Keyword Planner or Ahrefs.

Check Output and Iterate

Watch for hallucinated data. LLMs sometimes generate fake statistics, fabricated URLs, or invented tool names. Every factual claim in AI-generated SEO content needs a manual check before publishing.

Not every model handles long inputs the same way. If you paste full articles or large keyword lists, check before pasting long inputs.

GPT-5 and Claude Opus 4.6 support 400,000 and up to 1,000,000 tokens in beta (200,000 standard) respectively, so long-form inputs work fine. Smaller or free-tier models may truncate your input without warning.

Iteration beats perfection. Run a prompt, review the output, and refine. Ask the model to “make the title tag more specific” or “cut the generic opening,” just as effective prompting suggests.

Build a Prompt Library

Combine prompts into a sequence for best results. Start with a content brief to plan scope. Follow with a header structure prompt to lock the outline.

Then run title tag and meta description prompts for on-page elements. This three-step sequence covers most new content needs in about 15 minutes.

One common mistake is treating LLM output as final copy. The prompts handle structure and ideation. The human editor handles accuracy and brand consistency.

Save your best-performing prompts with customizations already filled in. A pre-filled template with your tone guidelines and audience description will produce consistent output from the first attempt. Building a personal prompt collection for marketing saves more time than any single prompt on its own.

Frequently Asked Questions

Stojan

Written by Stojan

Stojan is an SEO specialist and marketing strategist focused on scalable growth, content systems, and search visibility. He blends data, automation, and creative execution to drive measurable results. An AI enthusiast, he actively experiments with LLMs and automation to build smarter workflows and future-ready strategies.

View all articles

Table of Content

    SHARE