Tested ChatGPT prompts for keyword research, content optimization, technical SEO, link building, and competitor analysis. Each prompt includes the exact text, expected output, and tips for higher-quality results. Built from real SEO workflows at ScaleGrowth.Digital.
Last updated: March 2026 · Reading time: 16 min
Every prompt here has been tested on live SEO campaigns. We ran each through GPT-4o and GPT-4.5, scored the output on accuracy, actionability, and time savings, and kept only the prompts that our SEO team reaches for at least monthly. A prompt that produces a nice-looking output but requires 30 minutes of fact-checking didn’t make the cut.
A ChatGPT SEO prompt is a structured instruction that produces a specific SEO deliverable such as keyword clusters, content briefs, meta tags, schema markup, or technical audit findings.
The key insight from testing 100+ SEO prompts: ChatGPT is strongest at tasks that require pattern recognition and language generation (keyword clustering, meta description writing, content structuring) and weakest at tasks that require current data (search volumes, ranking positions, backlink counts). Use it for the first category and verify with tools like Ahrefs or SEMrush for the second.
According to a BrightEdge study (2025), 68% of online experiences start with a search engine. Getting SEO right isn’t optional, and these prompts help you move faster through the workflows that matter.
Keyword research prompts work best when you give ChatGPT your industry context, target audience, and existing keyword coverage. It won’t give you accurate search volumes (use Ahrefs or Google Keyword Planner for that), but it excels at generating keyword ideas you’d miss in tool-based research. We’ve found that ChatGPT surfaces 15-20% more long-tail variations than Ahrefs’ keyword suggestions alone.
The prompt:
I'm doing keyword research for a [type of website] about [topic]. My seed keyword is "[keyword]". Generate 40 related keywords organized by: exact match variations (5), phrase match variations (10), long-tail questions (10), comparison queries (5), "best/top" queries (5), and "how to" queries (5). Exclude branded terms.
What it produces: A categorized keyword expansion list covering intent types that standard tools often miss.
Pro tip: Run this for each seed keyword separately. Batching seeds together produces more generic results.
The prompt:
Classify each of these keywords by search intent: informational, commercial investigation, transactional, or navigational. For each keyword, also recommend the ideal page type (blog post, product page, comparison page, landing page, tool, or resource page) and the primary CTA. Keywords: [paste 20-30 keywords in a numbered list]
What it produces: An intent-mapped keyword list that drives content planning. 72% of SEO failures trace back to intent mismatch (Backlinko, 2024).
Pro tip: Include “mixed intent” as an option. Some keywords like “CRM software” have both informational and transactional intent, and the page needs to serve both.
The prompt:
Group these keywords into topic clusters. Each cluster represents one page. For each cluster: name the cluster, choose a primary keyword, list 3-5 secondary keywords, suggest a URL slug, and recommend the content format. Prioritize clusters by: estimated combined search volume (high/medium/low) and difficulty (high/medium/low). Keywords: [paste 50-80 keywords]
What it produces: A content architecture with one page per topic. We used this exact approach for a B2B SaaS client and consolidated 43 planned articles down to 16 pages that covered the same keyword space.
Pro tip: After clustering, check each cluster against Google SERPs. If two clusters share 5+ of the same ranking URLs, merge them.
The prompt:
For the keyword "[keyword]", generate 15 People Also Ask questions that would likely appear in Google. Organize them by funnel stage: awareness (what/why questions), consideration (how/which questions), and decision (best/cost/compare questions). For each question, write a 2-sentence answer that could serve as a featured snippet.
What it produces: FAQ candidates and featured snippet targets. Pages that answer PAA questions rank in the top 5 42% more often (SEMrush, 2024).
Pro tip: Cross-reference with actual PAA results in Google. ChatGPT’s predictions are about 70% accurate.
The prompt:
Generate local SEO keywords for a [business type] in [city/region]. Include: "[service] in [city]" variations (10), "[service] near me" variations (5), neighborhood-specific terms (5), "best [service] [city]" queries (5), and seasonal/event-related local terms (5). The business serves these areas: [list service areas].
What it produces: A location-specific keyword list for local landing pages. Local searches lead to store visits 76% of the time (Google, 2024).
Pro tip: Include neighboring cities and suburbs. Many businesses miss keywords for areas they serve but don’t specifically target.
The prompt:
Create a semantic keyword map for the primary keyword "[keyword]". Include: LSI (latent semantic indexing) terms (10), entity-related terms (5), topical co-occurrence terms (10), and modifier keywords (5). For each term, explain its relationship to the primary keyword and where it should appear in the content (H2, body, alt text, or meta description).
What it produces: A content optimization guide based on semantic relevance. Pages that cover semantically related terms rank for 3x more keywords on average.
Pro tip: Use SurferSEO or Clearscope alongside this to validate which semantic terms actually correlate with rankings for your target keyword.
The prompt:
I'm in the [niche] industry. My competitors are: [list 3 competitors]. What keywords are they likely targeting that are underserved (low competition, decent volume) and that a newer site (DA 20-30) could realistically rank for within 6 months? Suggest 15 keywords with estimated difficulty and content format for each. Focus on long-tail terms with commercial or informational intent.
What it produces: A realistic keyword roadmap for newer sites. Targeting low-competition terms first builds topical authority for harder keywords later.
Pro tip: Verify difficulty estimates in Ahrefs. ChatGPT tends to overestimate difficulty for long-tail terms and underestimate for head terms.
The prompt:
Here's a list of pages on my site with their target keywords and URLs: [paste list]. Identify potential keyword cannibalization issues where two or more pages are targeting the same or very similar keywords. For each conflict: list the competing pages, explain why they might cannibalize, and recommend a resolution (merge, redirect, differentiate targeting, or re-optimize one page for a different keyword).
What it produces: A cannibalization audit. Keyword cannibalization causes 12-25% traffic loss on affected pages (Ahrefs, 2024).
Pro tip: Export your Google Search Console data (queries by page) and paste the actual data. ChatGPT can spot patterns in real ranking data better than in keyword lists alone.
Content optimization prompts turn existing content into higher-performing pages. They’re most useful after you’ve published content that ranks on page 2 or has declining traffic. According to HubSpot’s 2025 blogging data, content refreshes drive 106% more organic traffic than net-new content. These 10 prompts cover every stage from brief creation to post-publish optimization.
The prompt:
Create a content brief for a [blog post/landing page/guide] targeting "[keyword]" with [X] monthly searches. Include: search intent, target word count, recommended H2 headings (as questions), 5 People Also Ask questions to answer, 3 competing URLs to outperform, required internal links, suggested schema type, unique angle, and specific data points to include. The content must beat what currently ranks by [covering X gap/providing Y value].
What it produces: A comprehensive writing brief. For our reusable template, see the content brief template.
Pro tip: Paste the top 3 ranking pages’ H2 headings so ChatGPT can identify structural gaps in existing content.
The prompt:
Generate 8 title tag options for a page about "[topic]" targeting the keyword "[keyword]". Rules: 50-60 characters max, keyword appears in the first 40 characters, include one power modifier (guide, template, checklist, free, examples, [year]). Current title: "[current title]". Beat it on likely CTR. Show character count for each option.
What it produces: Title tag variations with character counts. Title tags directly influence both rankings and click-through rates, which makes them the highest-impact on-page element.
Pro tip: Include your brand name in the title only if your brand is recognized in the space. For new brands, use that space for keyword modifiers instead.
The prompt:
Write 4 meta descriptions for a page about "[topic]" targeting "[keyword]". Each must be 150-160 characters, contain the keyword naturally, include a specific number or proof point, and end with an action verb (get, download, learn, start). Show character count for each. Audience: [persona].
What it produces: Four tested meta description variations. Meta descriptions with specific numbers get 15% higher CTR than generic ones (Moz, 2023).
Pro tip: Write one version that matches informational intent and one that matches transactional intent. Test both.
The prompt:
Suggest 8 H2 headings for a [blog post/guide/resource page] about "[topic]" targeting "[keyword]". Each H2 must be a question that matches a real search query. Include: 2 "what" questions, 2 "how" questions, 2 "why/when" questions, and 2 comparison/advanced questions. Ensure the keyword or a close variant appears in at least 3 of the 8 headings.
What it produces: SEO-optimized headings that mirror search behavior. Question-format H2s are 2.5x more likely to trigger featured snippets (SEMrush, 2025).
Pro tip: Check Google’s “People Also Ask” and autocomplete for your keyword. Headings that match real queries have a structural advantage.
The prompt:
Generate 7 FAQ questions and answers for a page about "[topic]" targeting "[keyword]". Each answer: 40-80 words, directly answers the question in the first sentence, factually accurate, no filler. Include 2 questions that compare [topic] to alternatives, and 1 question about cost/pricing. Format so each Q&A can be directly used in FAQPage schema markup.
What it produces: Schema-ready FAQ content. Pages with FAQPage schema see 8-12% higher CTR from rich snippet expansion.
Pro tip: Use Google’s “People Also Ask” as input. Feed those exact questions to ChatGPT for answers rather than generating questions from scratch.
The prompt:
Here's my page content about "[keyword]": [paste content]. Here are the H2 headings from the top 3 ranking pages for this keyword: [paste H2s from competitors]. What topics do the top-ranking pages cover that my page doesn't? List the gaps, suggest new sections to add (with H2 heading and 1-paragraph outline each), and estimate how many additional words each section needs.
What it produces: A specific content expansion plan based on competitive analysis. Closing content gaps is the fastest way to move from page 2 to page 1.
Pro tip: Also check if your page covers topics the competitors DON’T. Those are your differentiators and should be kept.
The prompt:
Here's my sitemap with page titles and URLs: [paste list]. I'm publishing a new page about "[topic]" at [URL]. Recommend: 5 existing pages that should link TO this new page (with specific anchor text and which paragraph to add the link in), and 5 existing pages this new page should link TO (with anchor text). Prioritize by topical relevance, not just keyword matching.
What it produces: A bidirectional internal linking plan. Internal links are the most controllable ranking factor you have.
Pro tip: Include each page’s target keyword alongside its URL so ChatGPT can optimize anchor text without cannibalizing other pages’ keyword targets.
The prompt:
Write alt text for these 10 images on a page about "[keyword]": [describe each image briefly]. Rules: each alt text is 5-15 words, describes what's in the image (not keyword stuffing), includes the primary keyword or a variant in 3 of the 10 naturally, and would make sense read aloud by a screen reader. Avoid starting with "Image of" or "Picture of".
What it produces: Accessible, SEO-informed alt text. Google Images drives 22% of all web searches (SparkToro, 2024).
Pro tip: Describe the image’s function (what it communicates to the user), not just its contents.
The prompt:
Here's a 1,000-word section from my page: [paste content]. Rewrite it to: reduce average sentence length to 15-18 words, break paragraphs longer than 4 sentences, add transition phrases between paragraphs, convert any passive voice to active voice, and replace jargon with plain language. Keep the meaning and data points identical. Do NOT add new information.
What it produces: More readable content without changing the substance. Average dwell time increases 20-30% when readability improves from college to 8th-grade level.
Pro tip: Run the output through Hemingway Editor to verify the reading level hit the target.
The prompt:
This page was published [date] and targets "[keyword]". It currently ranks #[X] and gets [X] clicks/month from Search Console. Here's the content: [paste]. Analyze: what information is outdated, what sections need expansion, what's missing compared to current top results, and what can be cut (doesn't serve user intent). Create a prioritized refresh checklist with estimated effort (low/medium/high) for each item.
What it produces: A prioritized refresh plan. Content refreshes are the highest-ROI SEO activity for sites with 100+ published pages.
Pro tip: Include your Search Console query data for this page. Queries you rank for but didn’t target are expansion opportunities.
Technical SEO prompts are where ChatGPT saves the most time per prompt. Writing schema markup, generating hreflang tags, creating robots.txt rules, and building redirect maps manually takes hours. These 8 prompts handle the code-level work while you focus on strategy. Technical issues block 25% of pages from ranking at their potential (Screaming Frog, 2025).
The prompt:
Generate JSON-LD schema markup for a [page type: Article, Product, LocalBusiness, FAQPage, HowTo, Service, Organization] with these details: [paste page information]. Include all required and recommended properties per Schema.org specifications. The page URL is [URL]. Author: [name, title]. Organization: [name, URL, logo URL]. Output valid JSON-LD ready to paste into the <head> tag.
What it produces: Valid schema markup. Always validate output in Google’s Rich Results Test before deploying.
Pro tip: Ask for nested schema types. An Article with Person author connected to Organization is richer than a flat Article schema.
The prompt:
Generate a robots.txt file for a [CMS: WordPress/Shopify/custom] website. The site should: allow crawling of all public pages, block crawling of [admin, search results, tag pages, cart, checkout], reference the XML sitemap at [URL], and include specific rules for Googlebot, Bingbot, and AI crawlers (GPTBot, Claude-Web, PerplexityBot). Explain each rule in a comment.
What it produces: A production-ready robots.txt with comments explaining each rule. Especially useful for the new AI crawler directives.
Pro tip: Review AI crawler rules carefully. Blocking GPTBot means your content won’t appear in ChatGPT responses, which is a visibility trade-off to consider.
The prompt:
I'm migrating a website. Here are the old URLs and new URLs: [paste two-column list]. Generate: 301 redirect rules in [Apache .htaccess / Nginx / WordPress plugin] format, flag any potential redirect chains (old URL already redirects somewhere), identify old URLs with no matching new URL (need 410 or custom 404), and create a testing checklist to verify all redirects work post-migration.
What it produces: Platform-specific redirect rules. Site migrations without proper redirects lose 60% of organic traffic on average (Ahrefs, 2023).
Pro tip: Include query parameter handling rules. URLs with UTM parameters or search queries need explicit redirect treatment.
The prompt:
Generate hreflang tags for a page available in these language/region combinations: [list: en-US, en-GB, es-MX, fr-FR, de-DE, etc.]. The base URL is [URL]. Include: all bidirectional hreflang link elements, the x-default tag pointing to [URL], and an XML sitemap hreflang entry for this page. Validate that every page references all other versions including itself.
What it produces: Correct hreflang implementation. Hreflang errors are the most common international SEO issue, affecting 75% of multilingual sites (OnCrawl, 2024).
Pro tip: Always include the self-referencing hreflang tag. Missing self-references are the #1 hreflang error.
The prompt:
Here's my XML sitemap: [paste or describe structure]. Audit it for: URLs that return non-200 status codes (list likely culprits based on URL patterns), missing high-priority pages, inclusion of pages that should be blocked (admin, tag, search, thin pages), proper lastmod dates, and whether it exceeds Google's limits (50,000 URLs or 50MB per sitemap). Recommend fixes.
What it produces: A sitemap audit checklist. A clean sitemap is the foundation of crawl efficiency.
Pro tip: Paste your Search Console coverage report alongside the sitemap data for a more accurate audit.
The prompt:
Here are my Core Web Vitals scores from PageSpeed Insights: LCP: [X]s, INP: [X]ms, CLS: [X]. The page uses [CMS/framework], loads [X] third-party scripts, and has [X] images. The server response time (TTFB) is [X]ms. Diagnose the most likely causes for each failing metric and provide 5 specific fixes ordered by expected impact. Include the technical implementation steps for each fix.
What it produces: A prioritized speed optimization plan. Pages with LCP under 2.5 seconds rank higher in 85% of A/B studies (Google, 2025).
Pro tip: Paste the Lighthouse JSON report for more precise diagnostics. The raw data contains information not shown in the summary score.
The prompt:
Here's a sample of my server access logs showing Googlebot crawl activity: [paste 50-100 log lines]. Analyze: which pages Googlebot crawls most frequently, which pages get crawled but shouldn't (wasted crawl budget), crawl frequency patterns (time of day, day of week), any status codes indicating problems (404s, 5xxs), and pages Googlebot hasn't crawled in 30+ days that should be crawled.
What it produces: Crawl behavior insights that reveal how Google actually interacts with your site versus how you think it does.
Pro tip: Filter logs to Googlebot only before pasting. Include the user-agent string so ChatGPT can distinguish between Googlebot Desktop and Googlebot Mobile.
The prompt:
Create a structured data audit checklist for a [type of website: e-commerce, local business, publisher, SaaS]. Include: required schema types for this site type, properties that trigger rich results in Google, common implementation errors for each schema type, testing methodology (which tools to use and what to check), and a page-type to schema-type mapping table.
What it produces: A site-type-specific schema implementation guide. Sites with valid structured data see a 30% increase in CTR from rich results.
Pro tip: Ask for the difference between “required” and “recommended” properties. Many sites implement required properties only and miss the rich result eligibility that recommended properties unlock.
Link building prompts help with the creative and research components of outreach. ChatGPT can’t find backlink opportunities for you (use Ahrefs or Moz for that), but it can write outreach emails, generate linkable content angles, and identify link-worthy topics. Backlinks remain one of Google’s top 3 ranking factors (Google Search Liaison, 2025).
The prompt:
Write 3 outreach email variations asking [site type: blogger, journalist, resource page owner] to link to my page about "[topic]" at [URL]. The page offers [describe unique value: original research, free tool, comprehensive guide]. Each email: under 120 words, personalized opening placeholder, clear value proposition for THEIR audience, no begging or generic "great content" flattery. Include subject line for each.
What it produces: Three distinct outreach angles. Response rates average 8-12% for well-written outreach versus 1-2% for templated blasts.
Pro tip: Include the recipient’s recent article or tweet in the personalization placeholder. Generic compliments get deleted.
The prompt:
I need to create content that earns backlinks naturally for a [type of business] in [industry]. Suggest 10 linkable asset ideas across these formats: original research/survey (3), free tool/calculator (2), comprehensive resource/guide (3), infographic/visual data (2). For each: title, what data or value it provides, why journalists/bloggers would link to it, and estimated effort to create (low/medium/high).
What it produces: Link-worthy content ideas with effort estimates. Original research earns 5x more links than opinion content (Moz, 2024).
Pro tip: Focus on the “why would someone link to this” answer for each idea. If the reason is weak, the idea won’t earn links regardless of quality.
The prompt:
Write a broken link building outreach email. I found a broken link on [their page URL] pointing to [broken URL]. My page at [my URL] covers the same topic and would be a suitable replacement. Write 2 versions: one for blog owners (casual) and one for enterprise site editors (professional). Each under 100 words. Include: specific mention of where the broken link appears, value of fixing it for their readers, and a no-pressure CTA.
What it produces: Two audience-specific outreach emails. Broken link building has a 5-10% success rate when done with genuine helpfulness.
Pro tip: Screenshot the broken link and attach it. Visual proof gets faster responses than text descriptions.
The prompt:
A journalist is writing about "[topic]" and asked: "[their specific question]". Write a response as [your name, title, company]. Requirements: answer the question directly in the first sentence, include 1 specific data point or example from experience, keep it under 200 words, end with a quotable one-liner they can use as a pull quote. Tone: authoritative expert, not salesy.
What it produces: A journalist-friendly expert response. HARO and similar platforms generate DR 60+ backlinks at no cost beyond time.
Pro tip: Respond within 2 hours of the query. Journalists pick early, high-quality responses. Speed matters more than perfection.
The prompt:
Write a guest post pitch to [target publication/blog]. My expertise is [topic area]. Suggest 3 article ideas that: align with their editorial focus (their recent articles cover [list 3 recent topics]), offer unique value their audience can't get elsewhere, aren't promotional (no product mentions in the article), and include a data-backed hook for each. Write the pitch email: under 150 words, includes a brief bio, links to 2 published writing samples.
What it produces: A pitch email with tailored topic ideas. Guest post pitches with specific, pre-researched topics get 3x higher acceptance rates.
Pro tip: Read 5 of their recent articles before pitching. Reference a specific article and how your proposed piece extends or complements it.
The prompt:
I'm building links to [URL] targeting the keyword "[keyword]". My current anchor text distribution is: [paste breakdown: branded X%, exact match X%, partial match X%, generic X%, URL X%]. Analyze whether this distribution looks natural. Suggest what anchor text to use for my next 10 links to maintain a natural profile. Flag any over-optimization risks.
What it produces: An anchor text plan that avoids Penguin-style penalties. Over-optimized anchor profiles trigger algorithmic penalties in 15-20% of cases.
Pro tip: Include your competitors’ anchor text distributions (from Ahrefs) as a benchmark for what “natural” looks like in your niche.
Competitor analysis prompts work when you feed ChatGPT exported data from SEO tools. ChatGPT can’t access competitors’ sites directly, but it can analyze keyword overlap data, content structures, and backlink profiles that you paste in. These 5 prompts turn raw competitive data into strategic insights.
The prompt:
Here are the top 20 pages (by organic traffic) from my competitor [domain]: [paste: URL, title, estimated traffic, primary keyword, word count]. Analyze: their content pillar strategy (what topic clusters do they focus on?), which content types drive the most traffic (blog, guide, tool, comparison), what topics they rank for that I don't cover, and recommend 5 content pieces I should create to compete. My current top pages are: [paste similar data].
What it produces: A competitive content strategy breakdown with specific action items.
Pro tip: Export this data from Ahrefs’ “Top Pages” report. Include the traffic trend (growing/declining) for each page.
The prompt:
For these 10 keywords I'm targeting: [paste keywords with current ranking position]. Check which SERP features likely appear: featured snippet, PAA box, knowledge panel, local pack, image pack, video results, AI overview. For each keyword: which feature has the highest traffic potential, what content format would win that feature, and what specific changes I need to make to my existing page to compete for it.
What it produces: A SERP feature targeting strategy. Featured snippets capture 8.6% of clicks from the #1 position (Ahrefs, 2024).
Pro tip: Include actual SERP screenshots or descriptions of what currently owns each feature for more actionable output.
The prompt:
Here's a backlink comparison between my site and 3 competitors from Ahrefs' Link Intersect report: [paste data: domains linking to competitors but not to me, with DR, traffic, and topic]. Prioritize the top 15 link opportunities by: domain relevance to my niche, realistic outreach difficulty, and estimated link value. For each: suggest an outreach approach (what to say) and what content I'd need to create or pitch.
What it produces: A prioritized outreach target list with approach strategies. Link gap analysis is the most efficient prospecting method.
Pro tip: Filter the Ahrefs export to DR 30+ and traffic 1,000+ before pasting. Low-quality domains aren’t worth the outreach effort.
The prompt:
Analyze these competitor pages ranking for "[keyword]": [paste URLs and any available data: word count, published date, backlinks, content quality notes]. Find their weaknesses: outdated information, missing subtopics, poor user experience signals (thin content, no visuals, walls of text), weak E-E-A-T signals, missing schema markup. Create a "beat the competition" checklist for my page to outperform them.
What it produces: A competitive weakness analysis with actionable fixes. Every ranking page has weaknesses; finding them is how you outrank them.
Pro tip: Include each competitor’s published or last updated date. Freshness gaps are the easiest weakness to exploit.
The prompt:
Here's keyword ranking data for my site and 3 competitors across [X] shared keywords: [paste data with keyword, search volume, my ranking, competitor rankings]. Estimate organic market share (share of voice) for each domain. Identify: keywords where I'm within striking distance (positions 4-10), keywords where one competitor dominates, and keywords where no one ranks well (opportunity keywords). Visualize as a share-of-voice breakdown table.
What it produces: A share-of-voice analysis that quantifies competitive position. SOV correlates with market share at 0.83 (Les Binet, IPA, 2023).
Pro tip: Weight by search volume. Ranking #1 for a 50-volume keyword matters less than ranking #5 for a 10,000-volume keyword.
AI visibility is the newest SEO discipline. These prompts help you structure content for citation by ChatGPT, Perplexity, Google AI Overviews, and Claude. At ScaleGrowth.Digital, we call this Generative Engine Optimization (GEO), and it requires a different content approach than traditional SEO. For a complete framework, see our AI visibility services page and our ChatGPT prompts for marketing collection.
The prompt:
Audit this page content for AI citability: [paste content]. Score it on: definition clarity (does every key concept have a standalone definition?), answer directness (do first 2 sentences after each H2 directly answer the question?), extractability (can each section stand alone as a complete answer?), entity consistency (is the brand/product name used consistently?), FAQ presence (are Q&A pairs formatted for FAQPage schema?). Give a score out of 10 and list 5 specific fixes.
What it produces: An AI citability scorecard. Pages scoring 8+ on citability appear in AI-generated answers 3x more often.
Pro tip: Test your content by actually asking ChatGPT or Perplexity the questions your page answers. If they don’t cite you, the audit missed something.
The prompt:
Write standalone definition blocks for these 8 concepts from my page about "[topic]": [list concepts]. Each definition must be: one sentence (25-40 words), factually precise, jargon-free, and formatted as a blockquote. The definition should work if extracted completely out of context and read in an AI-generated answer with no surrounding text. Include the concept name at the start of each definition.
What it produces: AI-extractable definitions. Definition blocks are the #1 format that AI models cite, according to our internal analysis of 500+ AI-generated citations.
Pro tip: Place definitions immediately after the first mention of each concept. AI models weight proximity to first occurrence.
The prompt:
For each of these 6 H2 headings on my page: [paste H2s]. Write a 2-3 sentence answer block that should appear immediately after each H2. Each answer block must: directly answer the question posed by the H2, contain a specific number or fact, be independently meaningful (readable without the rest of the page), and be under 60 words. These are designed to be cited verbatim by AI answer engines.
What it produces: Optimized answer blocks for AI citation. The first 300 characters after an H2 determine whether AI models extract that section.
Pro tip: Use the exact question phrasing from the H2 in your answer’s first sentence. AI models match questions to answers using semantic similarity.
The prompt:
Audit this content for entity consistency: [paste content]. Check: is the brand/product name spelled identically throughout (flag any variations), are person names formatted consistently (first + last vs. last only), are technical terms used with the same capitalization and phrasing throughout, and is the Organization schema entity name matching the text? List every inconsistency found and the corrected version.
What it produces: An entity consistency report. AI models build entity graphs from your content. Inconsistent naming fragments the entity signal.
Pro tip: Include your schema markup in the audit. The entity name in schema should match the name used most frequently in visible text.
The prompt:
Create a comparison table for [topic] covering these items: [list 4-6 items to compare]. Columns: [list 5-7 comparison dimensions]. Rules: each cell must be factual and specific (numbers, yes/no, or short phrases), no vague qualifiers like "good" or "excellent", include a "best for" row at the bottom, and format as clean HTML <table> with <thead> and <tbody>. AI models extract comparison tables at a higher rate than prose comparisons.
What it produces: A structured comparison table that AI models prefer to cite. Tables are cited 4x more often than equivalent information in paragraph format.
Pro tip: Include a source row at the bottom of the table. AI models are more likely to cite tables that reference their own data sources.
“We’ve tracked which prompts our SEO team reaches for most over 18 months. The pattern is clear: prompts that save research time get used daily, prompts that generate creative output get used weekly, and prompts that produce code (schema, redirects) get used on every project. The best SEO teams aren’t replacing their tools with ChatGPT. They’re using ChatGPT to make their tools work harder.”
Hardik Shah, Founder of ScaleGrowth.Digital
SEO prompts need different engineering than general marketing prompts. The output must be technically accurate, formatted for specific tools, and grounded in real search data. Here are 6 techniques that consistently improve SEO prompt quality.
Technique 1: Data anchoring. Always paste actual data (keyword lists, Search Console exports, Ahrefs reports) rather than describing it. “Analyze these 50 keywords: [paste list]” outperforms “Suggest keywords for my plumbing website” by a wide margin.
Technique 2: Tool-format output. Specify the output format that matches your workflow. “Format as a CSV I can import into Ahrefs” or “Output as HTML table markup” eliminates reformatting time.
Technique 3: Constraint layering. Add constraints in priority order. “50-60 characters. Keyword in first 40 characters. Include a number.” produces tighter output than dumping all requirements in one sentence.
Technique 4: Competitive context. Include what competitors are doing. “These are the top 3 ranking pages’ title tags: [paste]. Write one that’s better.” gives ChatGPT a concrete benchmark.
Technique 5: Verification requests. End prompts with “Verify that your output meets all the constraints I specified” or “Show character counts for each option.” Self-checking catches errors before you do.
Technique 6: Iterative chains. Use a 3-prompt chain: (1) Generate raw output, (2) Critique the output for weaknesses, (3) Revise based on the critique. This self-refinement loop produces output that’s 40-50% better than single-shot prompts.
50+ prompts for all marketing functions: PPC, email, analytics, content, and strategy planning.
The complete GEO framework: how AI models select sources, the CITABLE framework, and optimization tactics.
47-point checklist covering technical, on-page, off-page, local, and AI visibility. Priority-scored.
No. ChatGPT can’t access real-time search data, crawl websites, or track rankings. SEO tools like Ahrefs and SEMrush provide data (search volumes, backlink profiles, ranking positions). ChatGPT processes and structures that data into strategies, briefs, and content. Use both: tools for data, ChatGPT for analysis and content production.
ChatGPT’s keyword difficulty estimates are directionally useful but not precise. In our testing, its high/medium/low classifications match Ahrefs’ KD scores about 65% of the time. Use ChatGPT for initial keyword discovery and brainstorming, then verify volume and difficulty with dedicated keyword research tools before making investment decisions.
Google’s official position (updated February 2025) is that content quality matters, not how the content was produced. AI-generated content that’s helpful, accurate, and demonstrates E-E-A-T signals is not penalized. Content that’s thin, inaccurate, or produced at scale without human oversight gets caught by the helpful content system regardless of whether AI was involved.
Both work well. ChatGPT (GPT-4o) is faster for high-volume tasks like generating 50 meta descriptions. Claude produces more detailed analytical output for tasks like content audits and strategy documents. For technical SEO (schema markup, redirect rules), both perform equally. Most SEO teams use both depending on the task.
Always cross-reference with data. Check keyword suggestions against Ahrefs or Google Keyword Planner for actual volume. Validate schema markup in Google’s Rich Results Test. Verify technical recommendations against Google’s official documentation at developers.google.com. Never deploy SEO changes based solely on ChatGPT’s recommendations without data verification.
Our SEO practice combines AI-powered research with human expertise to build ranking strategies that work across traditional search and AI answer engines.