Hierarchical Site Summaries
WEEK 61 :: POST 1
Instructions Given to each A.I. - Please provide 3 prompt variations that share this objective:
Using AI to Analyze an Entire Website: Consolidating AI Chats
Overview: When faced with a mountain of pages, divide and conquer. Split your website content into logical chunks (by topic, category, or site section) and have the AI summarize each chunk separately. Once you have these individual summaries, feed them back into the AI (or a second-stage prompt) to generate a higher-level summary or to answer questions using the summaries as context. This hierarchical summarization approach yields a top-down understanding: first the details on each page, then the big picture of the entire site. It’s a scalable technique — whether your site has 10 pages or 1000, the AI deals with bite-sized pieces at each step. The end result is a concise but comprehensive overview that overcomes token limits by processing information in stages rather than all at once.
ChatGPT 5 :: Responses with Full Details :
A Quick Summary of the Three Fact vs. Opinion Prompt Variations
Beginner (Simple Map → Reduce): No-code, lightweight text/URL batches, fast rollups with a final Q&A index.
Intermediate (Taxonomy-First + Rubrics): Adds CSV structure, JSON outputs, quality scores, and action lists.
Advanced (Map → Calibrate → Reduce): Introduces calibration/QA, evidence matrices, JSONL artifacts, and strategy metrics.
ChatGPT Prompt Variation 1: One-Page-at-a-Time, Then Roll-Up
Introductory Hook
Big site, tiny context window? No problem. The divide-and-conquer approach turns an overwhelming website into bite-sized, AI-digestible chunks so you get accurate summaries (and answers) without hitting token limits.
Current Use
Hierarchical summarization matters because most models can’t read your whole website at once. By processing pages in small batches, then rolling them up into section summaries and one master synthesis, you get speed, coverage, and clarity—whether you’ve got 10 pages or 1,000.
Prompt: “Act as a hierarchical website summarizer. I will give you small batches of page text or URLs (≤15 pages per batch). For each page: (1) extract the main topic in ≤10 words, (2) produce a 120–180 word plain-English summary, (3) list 3–5 key bullets (facts, claims, data), and (4) capture metadata: {url, section/category, publish date if present}. Avoid speculation; if information is missing, write ‘NOT APPLICABLE.’ After I say ‘SECTION COMPLETE,’ generate a 250–350 word section summary that consolidates only the pages I’ve provided so far for that section. After all sections are complete and I say ‘SITE SYNTHESIS,’ produce a 400–600 word site-level summary plus a concise Q&A index of 10–15 likely questions with short answers based only on the summaries you created.”
Prompt Breakdown How A.I. Reads the Prompt:
["Act as a hierarchical website summarizer."] : Sets role and objective—AI prioritizes layered (page → section → site) outputs.
["I will give you small batches … (≤15 pages per batch)."] : Constrains scope to avoid token limits and encourages iterative runs.
["For each page: (1)…(4)"] : Standardizes page outputs, making later consolidation consistent and reliable.
["Avoid speculation; … ‘NOT APPLICABLE.’"] : Hallucination guard—you prefer missing data over invented facts.
["‘SECTION COMPLETE’ → 250–350 word section summary"] : Clear phase change from page-level to section-level map→reduce.
["‘SITE SYNTHESIS’ → 400–600 word site summary + Q&A index"] : Final reduce step and quick-answer surface for stakeholders.
Practical Examples from Different Industries
Tech Startup: Summarize product docs by module (Auth, Billing, Analytics) to brief new PMs in 30 minutes.
Retail eCommerce: Digest category pages (Shoes, Apparel, Accessories) to spot gaps and outdated copy.
Freelance Consultant: Scan a client’s blog by topic (SEO, Email, CRO) to prep a strategic content audit.
Creative Use Case Ideas
Pre-sales briefings for SDRs on a prospect’s site.
Rapid competitor comparison (run the same pipeline on 3 sites).
Investor prep: compress a portfolio company’s documentation by team.
Hiring: turn a careers site into a “who we are” cheat sheet.
Adaptability Tips
Change batch size (≤10 vs ≤25) depending on model/context.
Add a “Risk/Claims” bullet if you’re auditing compliance-sensitive sites.
Swap the Q&A index for an “Executive Talking Points” slide outline.
Optional Pro Tips
Ask for a “Terminology Glossary (acronym → definition)” at section level.
Add “Content Freshness Notes” to flag stale publish dates.
Request a “Duplicate/Overlap” detector to spot repeated content.
Prerequisites
A list of URLs or copied page text per batch.
A simple site sectioning plan (e.g., by URL path or menu).
Tags and Categories
Tags: summarization, website-audit, research, workflows
Categories: AI for Ops, Content Intelligence
Required Tools or Software
Any chat model (ChatGPT, Claude, Gemini). No code required.
Difficulty Level: Beginner
FAQ
Can I paste HTML? Yes, but stripping nav/boilerplate improves accuracy.
What if pages are long? Split per page into multiple excerpts and keep per-excerpt outputs.
Can I reuse summaries later? Yes—treat them as your site’s knowledge base.
Recommended Follow-Up Prompts
“Prioritize opportunities: From the section summaries, create a top-10 action list with impact/effort.”
“Generate an FAQ page using only the Q&A index you created.”
ChatGPT Prompt Variation 2: Section-Aware MapReduce with Rubrics
Prompt: “You are an AI assistant performing hierarchical summarization for a website using a taxonomy-first approach. I will provide (a) a CSV with {url, section, optional sub-section} and (b) page text batches (≤20 pages/batch). For each page, produce a JSON block: {url, section, sub_section, title_if_present, summary_150_220_words, key_points:[3–6], evidence_citations:[quoted lines or ‘NOT APPLICABLE’], freshness_note:[publish/update date if present else ‘NOT APPLICABLE’], quality_scores:{clarity:1–5, specificity:1–5, completeness:1–5}}. When I say ‘SECTION ROLLUP: <section>’, synthesize all JSON you created for that section into: {section_overview_300_450_words, themes:[3–7], gaps:[2–5], glossary:[term: definition], recommended_actions:[3–7 ranked]}. When I say ‘SITE SYNTHESIS,’ merge all section outputs into a 600–800 word executive summary + a table of contents (sections → one-liners). Use only supplied text; if unsure, write ‘NOT APPLICABLE.’”
Prompt Breakdown How A.I. Reads the Prompt:
["taxonomy-first approach"] : Groups pages by section to reduce cross-talk and produce cleaner rollups.
["CSV with {url, section,…}"] : A lightweight structure the AI can honor without code.
["JSON block per page … quality_scores"] : Machine-readable outputs enable filtering/sorting later.
["evidence_citations"] : Encourages quoting to reduce hallucinations and improve traceability.
["SECTION ROLLUP … {overview, themes, gaps, glossary, actions}"] : Standardized section deliverables for decision-making.
["SITE SYNTHESIS … executive summary + TOC"] : A navigation layer for stakeholders.
["Use only supplied text"] : Explicit guardrail against inventing content.
Practical Examples from Different Industries
SaaS B2B: Roll up Help Center + Product pages to map feature coverage and documentation gaps.
Healthcare Clinic: Summarize service pages by specialty; add glossary to translate jargon for marketing.
Nonprofit: Compress program pages by initiative; extract themes for grant proposals.
Creative Use Case Ideas
Build a “content quality dashboard” by averaging quality_scores per section.
Rapid SEO triage: themes → target clusters; gaps → content briefs.
Internal training packets for onboarding (section summaries → handouts).
Adaptability Tips
Expand quality_scores with “trustworthiness” or “update_recency.”
Switch JSON to CSV if your workflow favors spreadsheets.
Ask for “policy flags” for sensitive claims (finance/health/legal).
Optional Pro Tips
Request a “Duplicate Theme Detector” at section rollup time.
Add a “Content Lifecycle” label: {evergreen | seasonal | time-sensitive}.
Generate a “Cross-Link Suggestions” list from overlapping themes.
Prerequisites
CSV (url, section, optional sub-section) and text/URL batches.
Basic notion of site taxonomy (menu paths or categories).
Tags and Categories
Tags: mapreduce, JSON-outputs, evidence-quotes, QA
Categories: Site Intelligence, Content Ops
Required Tools or Software
Any chat model; spreadsheet app for CSV; optional Docs/Notion to store rollups.
Difficulty Level: Intermediate
FAQ
Do I need to crawl? Not necessarily—export URLs from your CMS or sitemap and paste text in batches.
What if a page belongs to multiple sections? Duplicate the entry with multiple section labels and let rollups deduplicate themes.
How do I use quality scores? Sort pages by low scores to prioritize edits.
Recommended Follow-Up Prompts
“From the gaps arrays, generate 5 content briefs per section with H2s and bullet outlines.”
“Create an internal glossary page from all section glossaries (dedupe terms).”
ChatGPT Prompt Variation 3: Two-Stage Synthesis with Guardrails & Metrics
Name: Map → Calibrate → Reduce (With Evaluation)
Reasonable Assumptions for Advanced Flow
Light automation allowed (optional script to fetch pages).
Token budgeting target: ≤1,500 words per page excerpt; ≤25 pages per batch.
Outputs stored as JSONL to reuse across models.
Prompt: “You are an advanced hierarchical summarization and evaluation assistant. Workflow:
Stage A – Page Summaries: I will provide batches of page text or cleaned HTML (≤25 pages/batch). For each page, output a JSONL line with: {url, section, title_if_any, page_summary_160_220_words, key_claims:[quoted_snippets], risks_or_ambiguities:[‘NOT APPLICABLE’ if none], metadata:{publish_or_update_date:‘value or NOT APPLICABLE’}}. Do not infer facts.
Stage B – Calibration (Quality & Coverage): After each batch, produce a calibration report with: {coverage_estimate:% of page content addressed, hallucination_risk:low/med/high with rationale, duplication_notes, terminology_list}.
Stage C – Section Reduce: When I say ‘SECTION REDUCE: <section>’, synthesize all JSONL for that section into: {section_summary_350_500_words, evidence_matrix:[{theme, supporting_pages:[urls]}], open_questions:[2–6], recommended_changes:[5–10 ranked]}.
Stage D – Site Synthesis: When I say ‘SITE REDUCE,’ merge all section outputs into: {exec_summary_700_900_words, strategic_findings:[5–10], cross-section_themes:[3–7], unanswered_questions:[3–7], suggested_dashboard_metrics:[5–8]}.
Rules: Use only provided text. Quote when citing. If data is missing, write ‘NOT APPLICABLE.’ Prefer precision over breadth. Keep JSON valid.”
Prompt Breakdown How A.I. Reads the Prompt:
["Stage A … JSONL line per page"] : Streaming-friendly artifact you can store/version.
["key_claims:[quoted_snippets]"] : Forces ground-truthing with verbatim citations.
["Stage B – Calibration"] : Adds quality control—surface coverage and hallucination risk early.
["Stage C – Section Reduce … evidence_matrix"] : Traceable claims → pages, ideal for audits.
["Stage D – Site Synthesis … dashboard_metrics"] : Turns summaries into measurable improvements.
["Rules … Use only provided text … Keep JSON valid"] : Strict guardrails to protect accuracy.
Practical Examples from Different Industries
Fintech: Use evidence_matrix to verify regulatory statements map to actual policy pages.
EdTech: Build cross-section themes (Curriculum, Pricing, Support) to align messaging across funnels.
Cybersecurity Vendor: Generate strategic_findings and unanswered_questions to guide roadmap and docs.
Creative Use Case Ideas
Build a private GPT grounded only in your JSONL summaries.
Create a changelog: rerun quarterly and diff section summaries.
Auto-generate an investor one-pager from exec_summary + metrics.
Adaptability Tips
Swap JSONL for Parquet/CSV based on storage.
Tighten token budgets (e.g., 120–160 words) for larger sites.
Add a “confidence_score 0–1” based on amount of quoted evidence.
Optional Pro Tips
Request a “Contradiction Scan” across sections to flag inconsistencies.
Add “Query Router Hints” (which section answers which question types).
Ask for a “Link Graph Sketch” (pages most referenced across sections).
Prerequisites
URL list, sitemap export, or cleaned HTML/page text.
Optional: lightweight fetcher (e.g., Python + Readability) to strip boilerplate.
Tags and Categories
Tags: evaluation, JSONL, evidence-matrix, governance
Categories: Content QA, Strategy & Insights
Required Tools or Software
Any LLM for summarization; optional local script to fetch pages; spreadsheet/DB or Notion to store outputs.
Difficulty Level: Expert
FAQ
What if some pages are paywalled? Exclude or include only accessible excerpts.
How do I ensure consistency across batches? Paste the same prompt each time; keep JSON schema identical.
Can I answer questions later from these summaries? Yes—treat JSONL as your RAG corpus.
Recommended Follow-Up Prompts
“From the evidence_matrix, produce a public-facing ‘What We Offer’ page with citations.”
“Create a quarterly review plan: which sections to re-audit and which metrics to track.”
Google Gemini 2.5 Pro :: Responses with Full Details :
A Quick Summary of How These Summarize & Validate Prompts Differ
Variation 1 (Beginner) focuses on atomic analysis, providing a straightforward prompt to summarize one single chunk of content at a time. It's designed for quickly understanding individual pages or sections in isolation.
Variation 2 (Intermediate) introduces a two-stage synthesis, guiding you to first create structured summaries for multiple pages and then feed those summaries back into a second prompt to generate a cohesive, high-level overview of the entire website.
Variation 3 (Expert) delivers an automated strategic report within a single, complex prompt. It uses custom tags to process all content chunks at once, performing both the individual analysis and the final synthesis to produce a comprehensive report complete with strategic insights and recommendations.
Gemini Prompt Variation 1: The Single-Chunk Summary
Introductory Hook: Are you staring at a digital mountain? Your website—or a competitor's—is packed with pages of valuable information, but extracting the core message feels like an impossible task. Manually reading through dozens of pages is a surefire way to lose a day. What if you could create a high-level overview in minutes, not hours?
Current Use: This beginner-friendly prompt is your first step into scalable content analysis. It’s designed to tackle one piece of the puzzle at a time, letting you summarize individual pages or content sections efficiently. By breaking a large website down into manageable chunks, you can conquer information overload and quickly grasp the key points of any page without getting lost in the details.
Prompt: "Assume the role of a meticulous Content Strategist. Your task is to summarize the following text, which is a single section from a website.
Your summary must be concise, under 150 words, and capture the following three elements:
1. **Core Purpose:** What is the primary goal of this text? (e.g., to sell a product, to inform, to provide support).
2. **Key Message:** What is the single most important takeaway for the reader?
3. **Target Audience:** Who is this text specifically written for?
Here is the text to summarize:
[PASTE WEBSITE PAGE/SECTION CONTENT HERE]"
Prompt Breakdown How A.I. Reads the Prompt:
["Assume the role of a meticulous Content Strategist."]: This instruction sets a specific persona for the AI. Instead of being a generic assistant, it adopts the mindset of a professional whose job is to analyze content for its strategic value, leading to a more focused and insightful summary.["Your task is to summarize the following text, which is a single section from a website."]: This clearly defines the AI's objective and scope. By specifying it's a "single section," you manage the AI's expectations and prevent it from looking for broader context that isn't there, improving the accuracy of the summary for that specific chunk.["Your summary must be concise, under 150 words..."]: This is a critical constraint. It forces the AI to be highly selective and distill the information down to its most essential elements, preventing long, rambling outputs and making the summaries easy to digest.["...and capture the following three elements: 1. Core Purpose, 2. Key Message, 3. Target Audience"]: This provides a structured output format. It directs the AI to look for specific types of information, moving beyond a simple summary to a strategic analysis. This structure ensures you get consistent, actionable insights from every piece of content.["Here is the text to summarize: [PASTE WEBSITE PAGE/SECTION CONTENT HERE]"]: This is the placeholder for your input. It clearly tells the AI where the source material begins, separating the instructions from the content to be processed.
Practical Examples from Different Industries
Tech Startup: A product manager pastes the content from their "Features" page. The AI returns a summary identifying the Core Purpose as "detailing technical capabilities," the Key Message as "our platform saves developers 50% of deployment time," and the Target Audience as "experienced software engineers and CTOs."
Small Retail Business: The owner of an e-commerce a store pastes the text from a product category page for "Handmade Leather Bags." The AI summarizes the Core Purpose as "driving sales through artisanal appeal," the Key Message as "our bags offer unique, durable style," and the Target Audience as "style-conscious consumers seeking quality craftsmanship."
Freelance Consultant: A marketing consultant pastes the text from their "Services" page. The AI identifies the Core Purpose as "generating client leads," the Key Message as "I provide data-driven SEO strategies to increase organic traffic," and the Target Audience as "small to medium-sized business owners."
Creative Use Case Ideas
Rapid Competitive Analysis: Quickly summarize the key pages of a competitor's website (Homepage, About Us, Product/Service pages) to understand their messaging and strategic positioning.
Content Refresh Audit: Use the prompt on your own blog posts to see if their original purpose and message are still clear. If the AI's summary is off-base, the content needs a rewrite.
Onboarding New Team Members: Summarize key internal documentation or sections of the company website to give new hires a high-level overview of different departments or products quickly.
Adaptability Tips
For Marketing: Modify the prompt to ask for a "customer-facing summary" and add a request for "Potential marketing angles" to generate ideas for ads or social media posts.
For Operations: Change the role to "Process Analyst" and use it to summarize internal SOPs (Standard Operating Procedures). Ask the AI to identify the "Main Objective" and "Critical Steps."
For Customer Support: Use the prompt to summarize knowledge base articles. Modify it to extract the "Problem Solved" and the "Final Resolution" to help train support agents.
Optional Pro Tips
Add
Tone of Voice:to the list of required elements to analyze the style of the content (e.g., "Professional," "Playful," "Technical").For longer texts, experiment with first summarizing with a higher word count (e.g., 300 words) and then using a second prompt to distill that summary further.
Specify the output format, such as
Please provide the summary in bullet points.for even faster reading.
Prerequisites
Access to the text content you want to summarize (e.g., by copying it from a webpage).
A clear idea of how you've chunked your website (e.g., by page, by section).
Tags and Categories
Tags: Summarization, Content Strategy, Website Audit, Competitive Analysis, Productivity.
Categories: AI for Entrepreneurs, Content Marketing, Productivity Hacks.
Required Tools or Software
Any standard generative AI tool (ChatGPT, Google Gemini, Anthropic Claude).
Difficulty Level
Beginner
Frequently Asked Questions (FAQ)
What's the ideal length for a "chunk" of text? Aim for the content of a single, focused webpage or a distinct section of a long page (e.g., 500-2000 words). If a page is extremely long, break it into logical sub-sections.
Will this work for any type of content? Yes, it's highly versatile. It works for blog posts, service pages, product descriptions, and "About Us" pages. The quality of the summary will depend on the clarity of the source text.
Recommended Follow-Up Prompts
"Based on the summary provided, generate 5 keywords that best represent this content."
"Rewrite the key message from this summary for a LinkedIn post."
Gemini Prompt Variation 2: The Two-Stage Hierarchical Summary
Introductory Hook: You've summarized individual pages, but you're still looking at a collection of puzzle pieces, not the full picture. How do you connect the dots between your product pages, your blog, and your company's mission? The next level of AI-driven analysis isn't just about summarizing—it's about synthesis.
Current Use: This intermediate prompt introduces a powerful two-stage workflow. First, you'll summarize your site's core sections in parallel. Then, you'll feed those individual summaries back into the AI to generate a single, cohesive "master summary." This hierarchical method mirrors how a human would read and comprehend a complex topic—by understanding the parts before assembling the whole. It’s the key to creating a true executive overview of an entire website.
Prompt:
Stage 1 Prompt (Run for each page/section):
"As a Senior Content Analyst, produce a detailed summary of the following text from a website. The summary should be structured as a 'Content Brief' with the following fields: - Page_Topic: [A 3-5 word title for the page's content] - Core_Objective: [What action or understanding is this page trying to drive?] - Key_Arguments: [List the 3 main points or benefits presented in a bulleted list] - Intended_Audience: [Describe the ideal reader in one sentence] Text to analyze: [PASTE WEBSITE PAGE/SECTION CONTENT HERE]"
Stage 2 Prompt (Run once, after collecting all Stage 1 summaries):
"You are a Chief Strategy Officer preparing an executive briefing. I will provide you with several 'Content Briefs,' each summarizing a different section of our company website. Your task is to synthesize these briefs into a single, high-level 'Website Strategy Overview.' This overview should be a 200-word narrative that answers: 1. What is the website's primary overall goal? 2. Who is the main audience it serves? 3. What is the core value proposition communicated across the entire site? Here are the content briefs: [PASTE ALL STAGE 1 SUMMARIES HERE]"
Prompt Breakdown How A.I. Reads the Prompt:
Stage 1 -
["As a Senior Content Analyst..."]: This role is more advanced than the first prompt, signaling to the AI that a more detailed, structured output is expected.Stage 1 -
["structured as a 'Content Brief' with the following fields..."]: This enforces a strict, machine-readable format. Using fields likePage_Topic:makes the output from Stage 1 consistent and perfect for input into Stage 2. It's a form of data normalization.Stage 2 -
["You are a Chief Strategy Officer preparing an executive briefing."]: This persona sets the altitude for the final output. The AI is no longer just summarizing; it's synthesizing information for a high-level strategic decision-maker, which means focusing on the big picture.Stage 2 -
["synthesize these briefs into a single, high-level 'Website Strategy Overview.'"]: The keyword here is synthesize, not summarize. This instructs the AI to find connections, patterns, and overarching themes between the individual briefs, rather than just restating them.Stage 2 -
["a 200-word narrative that answers..."]: This defines the final deliverable: a fluid, narrative report, not just a list of facts. The guiding questions ensure the narrative addresses the most critical strategic aspects of the website.
Practical Examples from Different Industries
Tech Startup: The team runs Stage 1 on their "API Documentation," "Pricing," and "Enterprise Solutions" pages. In Stage 2, the AI synthesizes these to report that the site's goal is to "drive enterprise adoption by showcasing robust, scalable, and well-documented API solutions for technical leaders."
Small Retail Business: The owner uses Stage 1 on "Men's Collection," "Women's Collection," and "Our Story" pages. The Stage 2 synthesis reveals the core value proposition is "providing ethically sourced, handcrafted apparel for eco-conscious millennials."
Freelance Consultant: A consultant runs Stage 1 on their "Blog," "Case Studies," and "Book a Call" pages. The Stage 2 overview concludes that the website's primary goal is to "establish authority through expert content and convert educated readers into high-value clients via demonstrated results."
Creative Use Case Ideas
Mergers & Acquisitions Due Diligence: Quickly get a strategic overview of a target company’s online presence and messaging.
Brand Repositioning: Analyze your current site to get a baseline understanding of your messaging, then use that insight to craft a new, more aligned strategy.
Investor Pitch Decks: Use the final "Website Strategy Overview" to create a concise slide in your pitch deck that explains your online market positioning.
Adaptability Tips
For Marketing: In Stage 2, change the persona to "Brand Manager" and ask for a "Brand Messaging Guide" that unifies the voice and tone across all sections.
For Operations: Use this process to synthesize summaries of different departmental SOPs into a company-wide "Operational Playbook."
For Product Development: Summarize user feedback from different channels (surveys, support tickets, reviews) in Stage 1. In Stage 2, ask the AI to synthesize a "Product Improvement Roadmap" based on recurring themes.
Optional Pro Tips
In Stage 1, add a field for
Keywords: [List the top 3-5 SEO keywords this page appears to target]. This data can be used in Stage 2 to analyze the overall SEO strategy.Experiment with different personas in Stage 2. For example, asking a "Venture Capitalist" persona to synthesize the briefs might yield an analysis focused on market viability and growth potential.
For a large number of briefs, instruct the AI in Stage 2:
First, group the content briefs by theme, and then write the synthesis.This helps the AI structure its thinking process.
Prerequisites
Willingness to perform a multi-step process.
The ability to copy and paste multiple pieces of text into a single prompt.
Tags and Categories
Tags: Hierarchical Summary, Synthesis, Content Strategy, Strategic Analysis, Website Audit.
Categories: AI for Entrepreneurs, Advanced Prompting, Business Strategy.
Required Tools or Software
Any standard generative AI tool (ChatGPT, Google Gemini, Anthropic Claude).
A text editor (like Notepad or Google Docs) to compile the Stage 1 outputs is helpful.
Difficulty Level
Intermediate
Frequently Asked Questions (FAQ)
How many Stage 1 summaries can I use in Stage 2? Most modern LLMs can handle a significant amount of text. You can typically paste 10-20 summaries (or more) without issue. For very large sites, you could do a three-stage process: summarize pages, then summarize categories, then summarize the entire site.
What if the Stage 1 summaries are inconsistent? The structured format of the Stage 1 prompt is designed to prevent this. By using fields like
Core_Objective:, you ensure every summary follows the same template, making the Stage 2 synthesis much more effective.
Recommended Follow-Up Prompts
"Based on the 'Website Strategy Overview,' write an 'About Us' page that reflects this strategy."
"Identify any potential gaps or inconsistencies in messaging based on the provided Content Briefs."
Gemini Prompt Variation 3: The Automated Hierarchical Analysis System
Introductory Hook: Manually copying, pasting, and synthesizing is powerful, but it isn't scalable. For the true AI trailblazer, the ultimate goal is automation. What if you could create a single, powerful prompt that acts as a self-contained system—processing an entire website's content in one go and delivering a multi-layered strategic analysis?
Current Use: This advanced prompt turns your AI into an automated analysis engine. It uses a sophisticated structure with clear delimiters to process multiple text chunks and perform a synthesis within a single interaction. You provide the raw content, formatted correctly, and the AI handles the entire hierarchical summarization process, from micro to macro. This is for entrepreneurs who want to build repeatable, scalable systems for intelligence gathering and content strategy.
Prompt: "You are an AI-powered Strategic Analysis System. Your primary directive is to perform a hierarchical content analysis based on the website data I provide. The process is sequential: first, analyze individual content chunks, and second, synthesize those analyses into a high-level executive summary.
I will provide the website content structured with the following tags:
<chunk title="[Page or Section Name]">
[Full text content of the page or section]
</chunk>
Your output must be a well-structured Markdown report containing two distinct sections:
### Section 1: Granular Content Analysis
For each chunk provided, you will generate a brief analysis containing:
- **Title:** The title I provided in the <chunk> tag.
- **Purpose:** A one-sentence summary of the content's primary goal.
- **Key Takeaways:** Three bullet points summarizing the most critical information for a reader.
### Section 2: Executive Synthesis
After analyzing all individual chunks, you will generate a holistic synthesis that includes:
- **Overarching Narrative (150 words):** A strategic narrative describing what the website, as a whole, communicates. It should identify the core value proposition and the journey it takes the visitor on.
- **Strategic Alignment Score (1-10):** Rate how well the individual chunks align with each other to support the overarching narrative. Provide a one-sentence justification for your score.
- **Identified Gaps & Opportunities:** Based on your analysis, list 2-3 potential content gaps, messaging inconsistencies, or strategic opportunities.
Begin analysis now. Here is the data:
[PASTE ALL YOUR <CHUNK> FORMATTED CONTENT HERE]"
Prompt Breakdown How A.I. Reads the Prompt:
["You are an AI-powered Strategic Analysis System..."]: This establishes the AI as an automated, logical entity, not just a creative assistant. It primes the AI for a structured, multi-step task.["I will provide the website content structured with the following tags: <chunk title="...">...</chunk>"]: This is the core of the innovation. You are teaching the AI a simple data format. This allows the AI to parse the input programmatically, separating and identifying each piece of content without confusion.["Your output must be a well-structured Markdown report..."]: This command ensures the output is not just a wall of text but a clean, readable, and easily shareable report using standard Markdown formatting (headings, bolding, bullet points).["### Section 1: Granular Content Analysis"]and["### Section 2: Executive Synthesis"]: This creates the hierarchical structure in the output itself. The AI is instructed to complete the first part (analysis of the pieces) before moving to the second part (synthesis of the whole), mirroring the human thought process.["Strategic Alignment Score (1-10)"]and["Identified Gaps & Opportunities"]: These are advanced analytical tasks. You are not just asking for a summary of what's there; you are asking the AI to perform a critical evaluation, find inconsistencies, and suggest strategic actions—providing immense value to an entrepreneur.
Practical Examples from Different Industries
Tech Startup: The founder pastes content from their Homepage, Pricing, and API docs, each in
<chunk>tags. The AI delivers a report. The Executive Synthesis gives a Strategic Alignment Score of 8/10, noting "strong alignment on developer-centric features but the homepage messaging is slightly too marketing-focused." It identifies an Opportunity to "create a clearer user journey from homepage discovery to API documentation."Small Retail Business: The owner inputs their "About Us," "Sustainability," and "Product" pages. The AI's report identifies a Gap: "The 'Sustainability' message is powerful but is not referenced on the Product pages, representing a missed opportunity to reinforce brand values at the point of sale."
Freelance Consultant: The consultant inputs pages for three different service lines (e.g., SEO, PPC, Content). The AI's synthesis might give a lower Alignment Score (e.g., 6/10) and justify it by stating, "Each service is presented as a silo. An overarching narrative that explains how these services integrate to deliver comprehensive client growth is missing."
Creative Use Case Ideas
Multi-source Research Synthesis: Instead of website chunks, use
<chunk title="[Source Name]">to paste articles, research papers, or reports on a single topic. The prompt will automatically create a literature review and synthesis.Team Brainstorm Consolidation: Have team members write their ideas on a topic. Place each person's contribution in a
<chunk>tag. The prompt will summarize each idea and then synthesize the group's overall direction and identify conflicting viewpoints.Drafting Annual Reports: Use
<chunk>tags for summaries from each department (Sales, Marketing, Ops). The AI will generate a cohesive draft of the company's annual narrative and highlight cross-departmental achievements.
Adaptability Tips
For Marketing: Add a field to the Executive Synthesis called
Proposed Slogan:that asks the AI to generate a tagline for the entire website based on its analysis.For Operations: Use this to analyze a series of project reports. In the synthesis section, ask for an
Overall Project Health AssessmentandCross-functional Risks.For Customer Support: Input transcripts of the top 10 most common support tickets. Modify the Executive Synthesis to identify
Root Cause Themesand suggest aProactive Support Strategy.
Optional Pro Tips
For ultimate automation with tools like ChatGPT's Code Interpreter, you can provide a
.zipfile with each text chunk as a separate.txtfile and modify the prompt to have the AI read, process, and analyze the files directly.Change the output format from Markdown to JSON by adding
Output the entire report as a single, valid JSON object.This makes the output machine-readable and allows it to be fed directly into other software, dashboards, or databases.Add a
Confidence Score (1-100%)to the analysis sections, asking the AI to rate its confidence in its own conclusions, which can help you gauge the reliability of the insights.
Prerequisites
Comfort with formatting text using simple tags.
The ability to gather all necessary text into a single document before running the prompt.
A robust LLM that can handle large context windows and complex instructions (e.g., GPT-4, Claude 3, Gemini Advanced).
Tags and Categories
Tags: Automated Analysis, System Prompting, Strategic Synthesis, Content Audit, AI Systems.
Categories: AI for Entrepreneurs, Advanced Prompting, Automation & Workflows.
Required Tools or Software
An advanced generative AI tool with a large context window (e.g., ChatGPT Plus, Gemini Advanced, Claude 3).
Difficulty Level
Expert
Frequently Asked Questions (FAQ)
Is the
<chunk>tag a real HTML or XML tag? No, it's a "pseudo-tag." We are simply using the format as a clear, logical delimiter that the AI can easily understand and parse. You could just as easily use[START CHUNK: Title]...[END CHUNK]. The key is consistency.What if my total text exceeds the AI's context window? This is the primary limitation of this advanced method. If you have a massive amount of text (e.g., a 500-page website), you would need to revert to the Intermediate method or run this prompt in batches (e.g., one batch for the "Marketing" section of the site, one for "Support," etc.).
Recommended Follow-Up Prompts
"Based on the 'Identified Gaps & Opportunities' in your report, develop a 3-month content plan to address these issues."
"Transform the 'Overarching Narrative' from the report into a 60-second elevator pitch for the company."
Claude 4.1 Sonnet :: Responses with Full Details :
A Quick Summary of How These Fact vs. Opinion Identifier Prompt Variations Differ
Variation 1 focuses on straightforward section-by-section summarization ("For each section I share, please create a 2-3 sentence summary..."). This beginner approach follows your existing website structure, processing predefined categories one at a time before creating a master summary.
Variation 2 emphasizes intelligent pattern recognition and topic discovery ("organize all provided content into 4-6 logical topic clusters based on thematic similarity, not site structure"). This intermediate approach lets AI find natural content relationships and identify gaps you might not have noticed.
Variation 3 deploys a sophisticated multi-agent system with specialized roles ("Agent 1: Content Extractor, Agent 2: Semantic Analyzer, Agent 3: Strategic Evaluator..."). This advanced approach creates recursive analysis loops, generates multi-stakeholder outputs, and adapts its analysis based on discovered patterns.
All three variations maintain the same fundamental goal: conquering website content overload through hierarchical summarization. They each break down large websites into manageable chunks, process them systematically, and build up to comprehensive overviews that overcome AI token limitations. The key difference lies in complexity and depth—from following existing structures (Variation 1), to discovering hidden patterns (Variation 2), to creating an intelligent analysis ecosystem (Variation 3). Choose based on your technical comfort level, time availability, and the depth of insights you need from your website content.
Claude.ai Prompt Variation 1: Simple Site Section Summarizer (Beginner)
Introductory Hook: Imagine trying to explain your entire company website to a potential investor in an elevator ride. You've got 30 seconds, but your site has 50+ pages of content. Sound familiar? Every entrepreneur faces this challenge: distilling complex web content into digestible insights without losing critical details. The solution isn't reading faster—it's reading smarter.
Current Use: This prompt transforms AI into your personal content analyst, breaking down website sections into clear, hierarchical summaries. Instead of drowning in information overload, you'll get structured insights that preserve both detail and context. Think of it as creating a "executive summary tree" where each branch represents a different section of your site, perfect for quick decision-making or strategic planning.
Prompt: "I need you to help me create a hierarchical summary of website content. First, I'll provide you with content from different sections of my website. For each section I share, please: 1) Create a 2-3 sentence summary capturing the main purpose and key information, 2) List 3-5 bullet points of essential details, 3) Note any unique selling points or critical data. After I've shared all sections, I'll ask you to create a master summary that synthesizes all the individual summaries into a cohesive overview of the entire website. Please confirm you understand and are ready to begin with the first section."
Prompt Breakdown How A.I. Reads the Prompt:
["I need you to help me create a hierarchical summary of website content"] : This establishes the overarching task and primes the AI to think in terms of levels and structure, not just flat summarization.
["First, I'll provide you with content from different sections of my website"] : Sets expectation for multi-turn interaction and tells the AI to prepare for chunked input rather than everything at once.
["For each section I share, please: 1) Create a 2-3 sentence summary... 2) List 3-5 bullet points... 3) Note any unique selling points"] : Provides a consistent framework the AI will apply to each chunk, ensuring uniform output structure across all sections.
["After I've shared all sections, I'll ask you to create a master summary"] : Signals a two-phase process and helps the AI understand it needs to retain information from earlier inputs for later synthesis.
["Please confirm you understand and are ready to begin"] : Creates a clear handshake protocol, ensuring the AI is aligned with the task before starting.
Practical Examples from Different Industries
Tech Startup Example: A SaaS company uses this to summarize their product documentation site, creating section summaries for Features, Pricing, API Docs, and Case Studies, then generating a master overview for sales teams to quickly reference during demos.
Small Retail Business: An online boutique applies this to their e-commerce site, summarizing product categories, shipping policies, and brand story pages to create training materials for new customer service representatives.
Freelance Consultant: A business strategist uses this to analyze competitor websites, breaking down their service offerings, methodology pages, and client testimonials into actionable competitive intelligence.
Creative Use Case Ideas
Audit your own website quarterly for content gaps and redundancies
Create onboarding documents for new team members from existing web content
Generate pitch deck content from your company website
Prepare for media interviews by summarizing your web presence
Build FAQ responses from existing site documentation
Adaptability Tips
For marketing: Add "identify emotional triggers and value propositions" to each section summary
For operations: Include "extract any processes or workflows mentioned" in the bullet points
For customer support: Modify to "highlight common questions this section might answer"
Optional Pro Tips
Feed section summaries back iteratively, asking for deeper analysis of high-priority sections
Use follow-up prompts to compare sections for consistency in messaging
Request summaries at different lengths (Twitter-length, paragraph, full page) for various uses
Prerequisites
Have your website content organized by logical sections or categories
Basic understanding of your site's information architecture
Ability to copy/paste or provide URLs for content sections
Tags and Categories Tags: #WebsiteAnalysis #ContentSummarization #HierarchicalProcessing #InformationArchitecture #ContentAudit Categories: Content Strategy, Website Optimization, Business Intelligence
Required Tools or Software
Any AI chat interface (ChatGPT, Claude, Gemini)
Access to your website content (can be live site or documents)
Optional: Web scraping tool for larger sites
Difficulty Level Beginner - Requires basic copy/paste skills and understanding of website sections
Frequently Asked Questions (FAQ) Q: How many sections can I process at once? A: Start with 5-7 sections per session to avoid token limits, then combine results.
Q: What if my site doesn't have clear sections? A: Group pages by purpose (sales, support, about) or user journey stages.
Q: Can this work with competitor sites? A: Yes, just ensure you're respecting copyright and terms of service.
Recommended Follow-Up Prompts
"Compare these summaries to identify content gaps"
"Generate a one-page executive brief from the master summary"
"Create a content improvement plan based on these summaries"
Claude.ai Prompt Variation 2: Smart Topic Clustering Summarizer (Intermediate)
For entrepreneurs who've had extensive back-and-forth with AI assistants on complex content projects, valuable insights often get buried in lengthy conversations. This intermediate-level prompt helps you systematically extract, organize and compile all meaningful content while maintaining a clear record of how the draft evolved through multiple iterations.
Prompt: "You are an expert content strategist performing a hierarchical website analysis. I will provide multiple pages of website content. Your task has three phases:
PHASE 1 - As I share content, identify and track recurring themes, topics, and keywords across all pages. Create a running list of topic clusters as you discover them.
PHASE 2 - Once I signal 'END OF CONTENT', organize all provided content into 4-6 logical topic clusters based on thematic similarity, not site structure. For each cluster, provide: a) Cluster name and theme, b) 3-4 sentence comprehensive summary, c) List of pages included, d) Key insights or patterns observed, e) Potential gaps or opportunities.
PHASE 3 - Create a master executive summary (200 words) that synthesizes all clusters into a cohesive narrative about the website's core value proposition and content strategy. Ready to begin analysis?"
Prompt Breakdown How A.I. Reads the Prompt:
["You are an expert content strategist performing a hierarchical website analysis"] : Establishes expertise persona, triggering more sophisticated analysis patterns and strategic thinking frameworks.
["Your task has three phases"] : Creates a clear mental model for the AI to follow, improving consistency and reducing confusion during multi-step processing.
["PHASE 1 - As I share content, identify and track recurring themes, topics, and keywords"] : Instructs AI to actively pattern-match and build a semantic map while processing, not just passively receive information.
["PHASE 2 - organize all provided content into 4-6 logical topic clusters based on thematic similarity, not site structure"] : Forces AI to think beyond surface-level organization and find deeper content relationships.
["For each cluster, provide: a) Cluster name... b) comprehensive summary... c) List of pages... d) Key insights... e) Potential gaps"] : Structured output requirements ensure comprehensive analysis with both descriptive and prescriptive elements.
["PHASE 3 - Create a master executive summary (200 words)"] : Specific word count forces conciseness while "cohesive narrative" encourages storytelling over bullet points.
Practical Examples from Different Industries
Tech Startup Example: A B2B software company discovers their content naturally clusters into "Integration Capabilities," "Security Features," "Customer Success Stories," and "Industry Solutions," revealing they're under-emphasizing integrations despite it being customers' top concern.
Small Retail Business: An artisan jewelry brand finds clusters around "Craftsmanship Process," "Sustainable Materials," "Custom Design Services," and "Gift Occasions," identifying an opportunity to create more content bridging sustainability with gift-giving.
Freelance Consultant: A marketing consultant's site reveals clusters of "Digital Transformation," "Brand Strategy," "Market Research," and "Training Programs," exposing that 70% of content focuses on strategy but 80% of revenue comes from training—a critical misalignment.
Creative Use Case Ideas
Reverse-engineer competitor content strategies by analyzing their topic clusters
Identify content collaboration opportunities by finding overlapping clusters between partner sites
Create dynamic navigation menus based on actual content clusters vs. assumed categories
Develop content calendars that maintain balance across all topic clusters
Generate keyword strategies based on naturally occurring topic relationships
Adaptability Tips
For SEO optimization: Add "identify primary and LSI keywords for each cluster"
For sales enablement: Include "map each cluster to buyer journey stage"
For content planning: Request "suggest 3 new content pieces for underrepresented clusters"
Optional Pro Tips
Run this analysis quarterly to track how your content focus shifts over time
Compare your topic clusters against industry leaders to identify differentiation opportunities
Use cluster analysis to inform site restructuring or navigation updates
Feed cluster summaries into AI to generate new content that bridges gaps between clusters
Prerequisites
Access to at least 10-15 website pages for meaningful clustering
Basic understanding of content themes in your industry
Ability to provide content in text format (HTML stripped preferred)
Tags and Categories Tags: #TopicModeling #ContentClustering #StrategicAnalysis #SEMAnalysis #ContentGaps Categories: Content Strategy, Information Architecture, Competitive Analysis
Required Tools or Software
AI platform with context window of 50K+ tokens (Claude, GPT-4, Gemini Pro)
Optional: Web scraping tools (Screaming Frog, ParseHub)
Optional: Text extraction tools for PDFs or complex pages
Difficulty Level Intermediate - Requires understanding of content strategy concepts and ability to interpret thematic analysis
Frequently Asked Questions (FAQ) Q: How do I know if my clusters are too broad or narrow? A: Aim for clusters that each represent 15-30% of your content; adjust the prompt for more or fewer clusters as needed.
Q: What if the AI creates overlapping clusters? A: Follow up with "Please consolidate overlapping clusters and reassign content to the most relevant single cluster."
Q: Can this identify SEO opportunities? A: Yes, gaps between clusters often represent untapped keyword opportunities.
Recommended Follow-Up Prompts
"Create a content gap analysis comparing these clusters to industry best practices"
"Generate 10 blog post ideas that would strengthen weak clusters"
"Develop a 90-day content strategy to balance cluster representation"
Claude.ai Prompt Variation 3: Dynamic Recursive Web Crawler & Synthesizer (Advanced)
Introductory Hook: Fortune 500 companies employ entire teams for comprehensive website audits that cost tens of thousands of dollars. What if you could replicate 80% of that value using an AI system that recursively crawls, analyzes, and synthesizes your web content like a team of analysts working in perfect coordination? This advanced prompt doesn't just summarize—it creates a living, breathing intelligence system for your web content.
Current Use: This sophisticated prompt chain transforms AI into a multi-agent analysis system, where each "agent" specializes in different aspects of content analysis before passing insights up the chain for synthesis. Entrepreneurs use this for deep competitive analysis, merger due diligence, comprehensive content audits, and strategic planning sessions where every detail matters but time is limited.
Prompt: "Initialize a multi-agent website analysis system with the following specialized agents and recursive processing protocol:
AGENT CONFIGURATION:
Agent 1 (Content Extractor): Process raw content, extract text, identify media elements, note interactive features
Agent 2 (Semantic Analyzer): Identify entities, concepts, sentiment, reading level, and create knowledge graphs
Agent 3 (Strategic Evaluator): Assess business value, user intent alignment, conversion optimization, competitive positioning
Agent 4 (Synthesis Director): Integrate all agent outputs into hierarchical summaries
PROCESSING PROTOCOL:
Level 1: For each URL/content provided, all agents analyze simultaneously, producing structured reports
Level 2: After every 5 pages, Synthesis Director creates intermediate cluster summary combining agent insights
Level 3: After all content, generate three-tier final output:
Tier A: Executive Dashboard (5 key metrics, 3 critical insights, 1 strategic recommendation)
Tier B: Department Briefs (500 words each for Marketing, Sales, Product, Customer Success)
Tier C: Technical Appendix (complete knowledge graph, entity relationships, content taxonomy)
RECURSIVE RULES:
If any agent identifies a critical insight, trigger deep-dive analysis on that specific area
When patterns repeat across 3+ pages, elevate to higher priority in synthesis
If contradiction detected, flag for human review with evidence from both sources
Begin with: 'System initialized. Agents ready. Provide first content for analysis.'"
Prompt Breakdown How A.I. Reads the Prompt:
["Initialize a multi-agent website analysis system"] : Triggers AI to compartmentalize different analytical functions, improving focus and reducing cognitive overload on any single task.
["AGENT CONFIGURATION: Agent 1... Agent 2... Agent 3... Agent 4"] : Creates distinct analytical personas, each with specialized focus areas, allowing for deeper expertise-based analysis.
["PROCESSING PROTOCOL: Level 1... Level 2... Level 3"] : Establishes a clear processing hierarchy that mirrors enterprise-grade analytical workflows, ensuring nothing is missed.
["Level 3: generate three-tier final output... Executive Dashboard... Department Briefs... Technical Appendix"] : Produces outputs tailored to different stakeholders and use cases from a single analysis run.
["RECURSIVE RULES: If any agent identifies... When patterns repeat... If contradiction detected"] : Implements intelligent decision trees that make the analysis adaptive and self-improving.
["Begin with: 'System initialized. Agents ready.'"] : Creates a clear state machine startup, ensuring AI understands the complex system is now active.
Practical Examples from Different Industries
Tech Startup Example: A fintech startup preparing for Series B uses this to analyze their entire platform documentation, generating investor-ready insights showing product maturity, market positioning, and technical differentiation. The executive dashboard becomes slide 3 of their pitch deck.
Small Retail Business: A growing e-commerce brand with 500+ products uses this to audit their entire catalog, discovering that 15% of products generate 78% of content engagement but only 34% of revenue, leading to a strategic pivot in merchandising.
Freelance Consultant: A digital transformation consultant runs this on Fortune 500 client websites before pitches, arriving with deeper insights than internal teams. One consultant landed a $250K contract by identifying digital inconsistencies the client hadn't noticed.
Creative Use Case Ideas
Pre-acquisition due diligence on target company websites
Real-time competitive intelligence gathering during product launches
Creating comprehensive knowledge bases from unstructured web content
Automated content compliance auditing for regulated industries
Building training datasets for custom AI models from web content
Generating multi-perspective SWOT analyses from single website crawls
Adaptability Tips
For International Markets: Add "Agent 5 (Cultural Analyzer): Assess localization, cultural sensitivity, regional relevance"
For E-commerce: Include "track product mentions, pricing patterns, inventory indicators" in Agent 1
For B2B Sales: Add "map content to Challenger Sale insights" to Agent 3
For Content Marketing: Have Agent 2 "identify content pillars and topic authority scores"
Optional Pro Tips
Chain multiple runs with different agent configurations for specialized analysis
Create "agent memories" by feeding previous analyses back as context
Use temperature settings of 0.3 for agents 1-3 (accuracy) and 0.7 for agent 4 (creativity)
Implement version control by saving each recursive analysis with timestamps
Build agent specialization over time by fine-tuning prompts based on output quality
Prerequisites
Familiarity with multi-step AI prompting and context management
Understanding of business analysis frameworks (SWOT, Porter's Five Forces)
Ability to preprocess and chunk large content sets
Experience interpreting multi-dimensional data outputs
Tags and Categories Tags: #EnterpriseAnalysis #MultiAgentSystems #RecursiveProcessing #CompetitiveIntelligence #AutomatedAuditing Categories: Advanced Analytics, Strategic Planning, Competitive Intelligence, Content Operations
Required Tools or Software
High-context AI model (Claude 3 Opus, GPT-4 Turbo, Gemini Ultra)
Web scraping tool with JavaScript rendering (Playwright, Puppeteer)
Optional: API access for programmatic implementation
Optional: Vector database for semantic search (Pinecone, Weaviate)
Difficulty Level Expert - Requires advanced prompt engineering skills, understanding of multi-agent systems, and ability to manage complex analytical outputs
Frequently Asked Questions (FAQ) Q: How do I manage token limits with this complex system? A: Process in batches of 10-15 pages, save intermediate outputs, then run final synthesis on saved summaries.
Q: Can agents really maintain their specialized roles? A: Yes, with clear role definition and consistent reinforcement through the recursive protocol.
Q: What's the optimal number of pages for meaningful analysis? A: 30-50 pages provides enough data for pattern recognition without overwhelming the system.
Q: How do I validate the accuracy of multi-agent outputs? A: Spot-check agent 1 extractions, verify 3-5 insights manually, and cross-reference strategic recommendations.
Recommended Follow-Up Prompts
"Create a 30-60-90 day action plan based on Tier A insights"
"Generate detailed competitor analysis using the same agent system"
"Develop predictive models based on patterns identified in Technical Appendix"
"Build custom agent for [specific business need] using this framework"
TAGS:
