How to Analyze Reviews with Claude AI: Export, Upload, and Extract Insights

Quick Summary: Export reviews from any of 11 platforms to CSV using Comment Exporter, upload the file to Claude, and run targeted prompts to extract sentiment patterns, feature requests, competitive intelligence, and actionable insights. Claude's 200K token context window lets you analyze 5,000+ reviews in a single conversation -- no batching required. This guide includes 10 copy-paste prompts that work immediately.

Why Claude AI for Review Analysis?

There are several large language models capable of analyzing customer reviews. ChatGPT, Gemini, and others all handle the task. But Claude, built by Anthropic, has specific architectural advantages that make it uniquely suited for review analysis at scale.

The headline advantage is Claude's 200K token context window. That translates to roughly 150,000 words of input in a single prompt. In practical terms, you can upload an entire CSV file containing 5,000+ reviews and analyze them in one conversation -- no splitting, no batching, no losing context between chunks. Most competing tools start struggling above 1,000 reviews. Claude handles five times that without breaking a sweat.

This matters because review analysis produces better results when the model can see the entire dataset at once. Patterns that only emerge at scale -- a complaint that appears in 3% of reviews, a feature request that clusters around a specific date range, a subtle sentiment shift between verified and unverified purchasers -- get lost when you split data across multiple conversations. Claude keeps the full picture in view.

Beyond context length, Claude brings several other strengths to review analysis. It follows complex instructions precisely, which means your structured prompts produce consistently formatted output. It excels at categorization tasks, sorting hundreds of individual complaints into coherent themes without dropping items or creating duplicate categories. It handles CSV parsing natively, reading column headers and maintaining the relationship between review text and metadata like star ratings, dates, and verification status. And it tends to hallucinate less than other models -- when Claude says 34% of reviews mention a specific complaint, that number is more likely to reflect the actual data.

Claude also offers an Artifacts feature that generates tables, charts, and formatted documents inline within the conversation. When you ask Claude to produce a sentiment breakdown table or a competitive comparison matrix, it renders them as structured visual elements you can copy, export, or reference -- not just plain text approximations.

If you have already read our ChatGPT review analysis guide, think of this as the companion piece. Same workflow philosophy -- export, upload, prompt -- but optimized for Claude's specific capabilities. The prompts in this guide are tuned to take advantage of Claude's strengths: large context processing, precise instruction following, and structured output generation.

What You Will Need

The entire workflow requires exactly two tools.

1. ✓Comment Exporter -- A Chrome extension that exports reviews and comments from 11 platforms (Amazon, Reddit, YouTube, Steam, Etsy, Product Hunt, Quora, Hacker News, Facebook, Google Maps, and Shopify) to CSV or JSON files. Reddit scraping is free. All 11 platforms are available on the All Access plan. Install it from the Chrome Web Store.
2. ✓Claude -- Anthropic's AI assistant. The free tier gives you access to Claude 3.5 Sonnet with file upload support and limited daily usage. The Pro plan ($20/month) unlocks Claude Opus 4, significantly higher usage limits, and the Projects feature for persistent analysis across conversations.

That is the complete toolset. No coding, no API keys, no data science background. If you can install a Chrome extension and upload a file, you can run every analysis in this guide.

Optional but useful: Google Sheets or Microsoft Excel for previewing and filtering your CSV files before uploading them to Claude. This step is not required but can help you focus your analysis on specific subsets of your data.

Step 1: Export Your Reviews with Comment Exporter

Claude cannot analyze reviews it cannot access. Platform review pages are locked behind pagination, infinite scroll, and dynamic loading. The first step is getting that data out of the platform and into a structured file that Claude can read.

The Comment Exporter Chrome extension handles this across all 11 supported platforms. Here is a quick walkthrough for the most common sources.

Exporting from Amazon

Navigate to any Amazon product page with reviews. Click the Comment Exporter icon in your browser toolbar and start the scrape. The extension automatically paginates through all review pages -- even products with thousands of reviews. Your exported CSV includes structured columns: review text, star rating, review title, date, verified purchase status, and helpful vote count. The extension supports 20+ Amazon country domains, so you can pull reviews from any market. See the full Amazon Review Scraper page for platform-specific details.

Exporting from YouTube

YouTube comments contain a different type of feedback -- shorter, more reactive, and often revealing about what content resonates with audiences. Navigate to any YouTube video, click the extension icon, and export. The CSV includes comment text, author, like count, timestamp, and reply status. The YouTube Comment Scraper handles videos with thousands of comments, including nested reply threads.

Exporting from Reddit

Reddit threads provide raw, unfiltered opinions. Comments are longer, more detailed, and often more honest than formal review platforms. Export any Reddit post to get comment text, author, score (upvotes minus downvotes), timestamp, and thread depth. Reddit scraping is free with Comment Exporter -- no Pro plan required. For specialized analysis techniques, see our Reddit sentiment analysis guide.

Other Supported Platforms

The same one-click export workflow works on Steam (game reviews with playtime data), Etsy (handmade and vintage product reviews), Product Hunt (launch feedback), Quora (answers and discussions), and Hacker News (tech community threads). One extension, 11 platforms, same CSV output format.

"Compared to other tools, the price is incredibly low for the quality it delivers. I can now export reviews from Amazon and Etsy in seconds." -- Mitran Marian

Step 2: Prepare Your Data for Claude

Claude handles raw CSV files well, but a few minutes of preparation can significantly improve the quality of your analysis output.

CSV Is the Best Format for Claude

Comment Exporter supports both CSV and JSON export. For Claude review analysis, CSV is the recommended format in almost every case. Claude parses tabular CSV data natively, reads column headers to understand data structure, and maintains relationships between review text and metadata columns throughout the analysis. CSV files are also smaller than their JSON equivalents, which means you can fit more reviews within the context window.

The one exception: use JSON when you need to preserve hierarchical thread structure, such as Reddit comment trees with parent-child relationships. For all other platforms -- Amazon, YouTube, Steam, Etsy -- CSV is the right choice.

What Columns to Include

Comment Exporter's CSV output includes all available metadata by default. For most Claude analyses, you will want to keep these columns:

• ✓Review text / comment body -- The core content for analysis. Always include this.
• ✓Star rating -- Essential for sentiment analysis and for filtering (analyzing 1-star reviews separately from 5-star reviews).
• ✓Date -- Required for trend detection and temporal analysis.
• ✓Verified purchase status -- Allows Claude to compare verified vs. unverified reviewer sentiment.
• ✓Helpful votes / like count -- Lets Claude weight high-signal reviews more heavily.

You can remove columns like author name or profile URL if they are not relevant to your analysis. Fewer columns means less token usage per review, which means you can fit more reviews in a single upload.

File Size and Review Count Tips

Claude's 200K token context window is generous, but here are practical guidelines:

• ✓Under 5,000 reviews: Upload the entire CSV in one file. Claude handles this comfortably in a single conversation.
• ✓5,000-10,000 reviews: Consider removing unnecessary columns to reduce token usage. If reviews are short (1-2 sentences each), you can often fit the full dataset. If reviews are longer (3+ paragraphs), consider splitting into two uploads.
• ✓10,000+ reviews: Pre-filter in Google Sheets or Excel before uploading. Filter by star rating, date range, or helpful vote count to create targeted subsets. A focused dataset of 3,000 relevant reviews produces sharper insights than 10,000 unfiltered ones.

For most real-world analysis scenarios, your exported CSV will be well within Claude's capacity. This is the key advantage -- where other tools force you to batch and split, Claude lets you analyze the full dataset in one pass.

"The files it spits out are perfectly formatted, no messy cleanup needed." -- Alfon Labadan

Step 3: Upload and Analyze in Claude

Getting your data into Claude is straightforward. There are two methods, and a powerful advanced option for ongoing analysis.

Method A: Direct File Upload (Recommended)

Both free and Pro Claude users can upload files directly into a conversation. This is the recommended approach for any dataset over 50 reviews.

1. ✓Open claude.ai and start a new conversation.
2. ✓Click the attachment icon (paperclip) in the message input area.
3. ✓Select your exported CSV file from Comment Exporter.
4. ✓Type your analysis prompt (see the 10 prompts below) in the message field.
5. ✓Hit send. Claude will confirm it has read the file, display the number of rows and columns detected, and begin analysis.

Claude reads the entire file contents -- including column headers and all data rows -- and keeps them in context for the duration of the conversation. You can ask follow-up questions, run additional prompts, and drill into specific findings without re-uploading the file.

Method B: Paste Directly

For smaller datasets -- under 100 reviews -- you can paste CSV content directly into the chat window. Open your CSV in a text editor, select all content, and paste it with your prompt.

Here are [NUMBER] reviews exported from [PLATFORM] for [PRODUCT NAME]. Analyze them according to the instructions below.

---

[PASTE CSV DATA HERE]

---

[YOUR ANALYSIS PROMPT HERE]

Using Projects for Persistent Analysis

Claude Pro users have access to the Projects feature, which is a significant advantage for ongoing review analysis. Projects let you pin files and instructions that persist across every conversation within that project.

Here is how to set up a review analysis project:

1. ✓Create a new Project in Claude (e.g., "Amazon Product Analysis" or "Q1 2026 Customer Feedback").
2. ✓Pin your exported CSV file to the project's knowledge base.
3. ✓Add a project-level instruction like: "You are a customer insights analyst. The attached CSV contains product reviews. When analyzing, always provide specific numbers, direct quotes, and structured output with clear headers."
4. ✓Start conversations within the project. Every conversation will have automatic access to the pinned CSV and will follow your standing instructions.

This is powerful for iterative analysis. You can run Prompt 1 (Sentiment Overview) in one conversation, Prompt 4 (Competitor Comparison) in another, and Prompt 7 (Product Improvement Priorities) in a third -- all within the same project, all referencing the same data, without re-uploading each time. You can also update the pinned CSV with fresh review exports as new data comes in.

10 Copy-Paste Claude Prompts for Review Analysis

These are the core of this guide. Each prompt is designed for a specific analysis use case, tuned for Claude's strengths, and ready to paste alongside your uploaded CSV file. Replace the placeholders in brackets with your actual values.

Every prompt works with reviews from any of the 11 platforms supported by Comment Exporter. For a broader look at AI-powered review analysis tools and techniques, see our guide to the best AI tools for review analysis.

Prompt 1: Sentiment Overview

Use case: Get a structured breakdown of overall customer sentiment. This is the best starting point for any review analysis project -- run this first, then drill deeper with the specialized prompts below.

You are a customer insights analyst. I have uploaded a CSV file containing [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Analyze every review in the file and produce a structured sentiment report:

1. OVERALL SENTIMENT BREAKDOWN
   - Percentage and count of Positive, Neutral, and Negative reviews.
   - If star ratings are available, show the distribution (1-star through 5-star).

2. TOP 5 THINGS CUSTOMERS LOVE
   For each theme:
   - Theme name
   - Percentage of reviews mentioning it
   - 3 direct quote examples from different reviewers

3. TOP 5 COMPLAINTS
   For each complaint:
   - Complaint description
   - Percentage of reviews mentioning it
   - 3 direct quote examples from different reviewers
   - Severity rating (Critical / Major / Minor)

4. SENTIMENT TREND OVER TIME
   If dates are available, break sentiment down by month or quarter. Note whether overall satisfaction is improving, declining, or stable.

5. EXECUTIVE SUMMARY
   Write a single paragraph (5-6 sentences) summarizing the key findings as if briefing a product manager who has 30 seconds to read it.

Be precise. Use actual counts and percentages from the data. Do not generalize or approximate when exact numbers are available.

Prompt 2: Top Complaints Extraction

Use case: Deep-dive into what is going wrong. Best used on filtered 1-3 star reviews, but works on full datasets too. Pairs well with our e-commerce review analysis guide for product-focused analysis.

You are a customer experience analyst. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Extract and categorize every distinct complaint, frustration, and negative experience mentioned across all reviews. For each complaint:

1. COMPLAINT: Describe the issue in one clear sentence.
2. CATEGORY: Assign to a category (e.g., Product Quality, Shipping, Customer Service, Price/Value, Durability, Ease of Use, Accuracy of Description, Packaging).
3. FREQUENCY: Exact count and percentage of reviews mentioning this complaint.
4. SEVERITY: Critical (product fails entirely), Major (significantly degrades experience), or Minor (annoyance but not dealbreaker).
5. CUSTOMER RESPONSE: What actions did complainants take? (returned product, contacted support, left warning for others, switched to competitor -- note which competitor if named)
6. REPRESENTATIVE QUOTES: 3 verbatim quotes from different reviewers.

After listing all complaints, provide:
- A PRIORITY RANKING of the top 10 complaints sorted by (frequency x severity).
- A CATEGORY SUMMARY showing which complaint categories have the most issues.
- THREE IMMEDIATE ACTIONS the company should take based on this data.

Prompt 3: Feature Request Mining

Use case: Build a data-driven product roadmap based on what customers actually ask for, wish existed, or cite as reasons for switching to competitors.

You are a product researcher. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Extract every feature request, improvement suggestion, and wishlist item mentioned by reviewers. Include both explicit requests ("I wish it had...") and implicit ones (complaints that imply a missing feature).

For each feature request:

1. FEATURE/IMPROVEMENT: Describe it clearly in one sentence.
2. CATEGORY: Group into a product area (e.g., Performance, Design, Software, Pricing, Compatibility, Accessories).
3. FREQUENCY: How many reviews mention this (exact count and percentage).
4. URGENCY SIGNAL: How do reviewers frame this?
   - Dealbreaker: "I returned it because..."
   - Strong want: "The only thing missing is..."
   - Nice-to-have: "It would be cool if..."
   - Competitive gap: "I switched to [competitor] because they have..."
5. REPRESENTATIVE QUOTES: 2-3 verbatim quotes from different reviewers.

At the end, provide:
- CATEGORY SUMMARY: Total feature requests per category.
- PRIORITY MATRIX: Top 10 features ranked by (frequency x urgency).
- COMPETITIVE FEATURE GAPS: Features that customers say competitors already offer.

Prompt 4: Competitor Comparison from Reviews

Use case: Understand competitive positioning based on real customer language. Works best when you upload reviews from 2-3 competing products in the same conversation. For a comprehensive framework, see our competitor analysis using reviews guide.

I have uploaded review data for competing products in the [CATEGORY] space:
- Product A: [NAME] ([NUMBER] reviews from [PLATFORM])
- Product B: [NAME] ([NUMBER] reviews from [PLATFORM])
- Product C: [NAME] ([NUMBER] reviews from [PLATFORM])

Analyze all reviews and produce a competitive intelligence report:

1. COMPETITIVE STRENGTHS
   For each product, list what customers say it does better than alternatives. Be specific -- not "good quality" but "battery lasts 2x longer according to 34% of reviewers."

2. COMPETITIVE WEAKNESSES
   For each product, list where customers say it falls short relative to competitors.

3. HEAD-TO-HEAD COMPARISONS
   When reviewers explicitly compare products, capture those comparisons verbatim. Which product wins on which dimensions?

4. SWITCHING TRIGGERS
   What specific reasons do customers give for leaving one product for another? List the trigger, the direction of the switch (from Product A to Product B), and example quotes.

5. UNMET NEEDS
   What do customers across ALL products complain about? These are category-level gaps that no current product addresses well.

6. POSITIONING SUMMARY TABLE
   Create a comparison table with products as columns and the top 8 customer-valued dimensions as rows. Rate each product as Strong / Average / Weak on each dimension based on review evidence.

Support every claim with direct quotes and review counts.

Prompt 5: Customer Persona Identification

Use case: Discover who your customers actually are based on how they describe themselves, their use cases, and their priorities in reviews. This is market research extracted from existing data.

You are a market researcher. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Analyze the reviews to identify distinct customer personas -- groups of reviewers who share similar characteristics, needs, and use cases. For each persona:

1. PERSONA NAME: Give a descriptive label (e.g., "Budget-Conscious Parent," "Power User Professional," "First-Time Buyer").
2. ESTIMATED SIZE: What percentage of reviewers fit this persona?
3. DEMOGRAPHICS SIGNALS: Any age, profession, experience level, or lifestyle indicators mentioned in reviews.
4. PRIMARY USE CASE: How does this persona use the product? What problem are they solving?
5. PURCHASE MOTIVATION: Why did they choose this specific product?
6. SATISFACTION LEVEL: What is the average sentiment for this persona? Are they generally happy or disappointed?
7. KEY COMPLAINTS: What does this persona specifically complain about?
8. KEY PRAISE: What does this persona specifically love?
9. REPRESENTATIVE QUOTES: 3-4 verbatim quotes that exemplify this persona's perspective.

Identify 4-6 distinct personas. After describing each, provide:
- A COMPARISON TABLE showing all personas side by side.
- MARKETING IMPLICATIONS: How should messaging differ for each persona?
- PRODUCT IMPLICATIONS: Which product improvements would matter most to each persona?

Prompt 6: Review Trend Analysis Over Time

Use case: Detect whether product quality, customer satisfaction, or specific issues are improving or declining. Requires review data with dates. Particularly valuable after product updates, pricing changes, or competitor launches.

You are a trends analyst. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME]. These reviews span from [START DATE] to [END DATE].

Analyze how customer feedback has changed over time. Produce a comprehensive trend report:

1. SENTIMENT TRAJECTORY
   Break down average sentiment by month or quarter. Is overall satisfaction improving, declining, or stable? Quantify the change.

2. RATING DRIFT
   If star ratings are included, calculate the average rating per month/quarter. Identify any significant shifts (0.3+ stars in either direction) and correlate with review content to explain why.

3. EMERGING ISSUES
   What complaints appear in recent reviews (last 3 months) that were absent or rare in older reviews? These may signal product changes, quality control problems, or new competitor pressure.

4. RESOLVED ISSUES
   What complaints appeared frequently in older reviews but have decreased or disappeared in recent ones? These may indicate successful product improvements or fixes.

5. SEASONAL PATTERNS
   Are there recurring patterns tied to time of year? (e.g., more shipping complaints during holidays, more returns in January, different use cases in summer vs. winter)

6. VOLUME TRENDS
   Is the review volume increasing or decreasing over time? Sudden spikes may indicate marketing campaigns, viral moments, or product issues.

7. KEY INFLECTION POINTS
   Flag specific months where sentiment shifted dramatically. For each, identify the likely cause based on review content.

Present all findings chronologically with specific date ranges, counts, and percentages.

Prompt 7: Product Improvement Priorities

Use case: Transform review data into an actionable improvement roadmap with clear priorities. Designed for product managers, founders, and anyone who needs to decide what to fix or build next.

You are a product strategist. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Analyze all reviews and produce a prioritized product improvement roadmap:

1. CATEGORY MAP
   Group all feedback into 6-8 product categories (e.g., Build Quality, Ease of Use, Customer Support, Price/Value, Performance, Documentation, Packaging, Compatibility). For each category, count the total mentions (both positive and negative).

2. SATISFACTION SCORE PER CATEGORY
   For each category, calculate the ratio of positive to negative mentions. Rate as:
   - High Satisfaction (>75% positive)
   - Medium Satisfaction (50-75% positive)
   - Low Satisfaction (<50% positive)

3. IMPROVEMENT OPPORTUNITIES
   For each Medium and Low satisfaction category, list specific improvements based on customer feedback. Each improvement should be concrete and actionable -- not "make it better" but "add USB-C charging port (requested by 23% of reviewers)."

4. IMPACT vs EFFORT MATRIX
   Classify each improvement as:
   - Quick Win: high customer impact, likely low implementation effort
   - Strategic Investment: high customer impact, likely significant effort
   - Fill-In: lower customer impact, low effort
   - Deprioritize: lower customer impact, high effort

5. RECOMMENDED ROADMAP
   List the top 8 improvements in the order they should be tackled. For each, provide:
   - The improvement
   - Why it is prioritized at this position
   - Expected customer impact (based on review frequency and severity)
   - Supporting quotes from reviews

Ground every recommendation in specific review data. No speculation.

Prompt 8: Marketing Language Extraction

Use case: Mine the exact words and phrases customers use to describe their problems, outcomes, and experiences. This is raw material for ad copy, landing pages, email sequences, and social media content -- language that resonates because it comes from the audience itself.

You are a copywriting researcher. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME].

Mine these reviews for customer language that can be used in marketing materials. Extract verbatim phrases -- do not paraphrase or clean them up.

1. PROBLEM LANGUAGE (Before Purchase)
   How do customers describe the problem they had BEFORE using this product?
   Extract 15-20 exact phrases. (e.g., "I was tired of..." / "Nothing else worked for..." / "I had been struggling with...")

2. OUTCOME LANGUAGE (After Purchase)
   How do customers describe the result AFTER using this product?
   Extract 15-20 exact phrases. (e.g., "Finally I can..." / "This changed how I..." / "I never thought I would be able to...")

3. OBJECTION LANGUAGE (Pre-Purchase Hesitation)
   What hesitations did customers mention before buying?
   Extract 10-15 exact phrases. (e.g., "I was skeptical about..." / "I almost did not buy because..." / "I was worried it would...")

4. COMPARISON LANGUAGE
   How do customers compare this to alternatives?
   Extract 10-15 exact phrases with the competitor or alternative named. (e.g., "Way better than [competitor]..." / "I switched from [alternative] because...")

5. EMOTIONAL TRIGGERS
   What emotions come through most strongly? List the top 5 emotions with frequency and 3 example quotes each.

6. MOST POWERFUL REVIEW EXCERPTS
   Select the 10 most compelling, quotable review snippets that could work as testimonials or social proof. For each, note why it is effective (specificity, emotion, relatability, outcome).

7. HEADLINE CANDIDATES
   Based on the most common and emotionally charged customer language, suggest 15 potential headlines or taglines that use the customers' own words.

Prompt 9: Verified vs. Unverified Review Comparison

Use case: Determine whether verified and unverified reviews tell different stories. Useful for detecting review manipulation, understanding bias, and deciding how much weight to give different review segments. Works best with Amazon reviews where verification status is included in the export.

You are a review quality analyst. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME]. The data includes a verified purchase indicator.

Compare verified and unverified reviews across every dimension:

1. VOLUME SPLIT
   How many reviews are verified vs. unverified? What percentage of total?

2. RATING COMPARISON
   What is the average star rating for verified reviews vs. unverified reviews? Show the full distribution (1-5 stars) for each group.

3. SENTIMENT COMPARISON
   Is sentiment more positive, negative, or neutral in one group vs. the other? Quantify the difference.

4. TOPIC DIFFERENCES
   What do verified reviewers talk about that unverified reviewers do not, and vice versa? Are there themes unique to each group?

5. LANGUAGE QUALITY
   Compare the average review length, specificity, and detail level between groups. Do verified reviews tend to include more product-specific details?

6. SUSPICIOUS PATTERNS
   Flag any signs of review manipulation in either group:
   - Clusters of reviews posted on the same date
   - Generic or repetitive language across multiple reviews
   - Reviews that read like marketing copy
   - Unusual rating patterns (e.g., 90% 5-star in unverified but 60% 5-star in verified)

7. RELIABILITY ASSESSMENT
   Based on this analysis, provide a data quality score (1-10) for each group and an overall recommendation on how much weight to give each group in decision-making.

8. KEY TAKEAWAYS
   List 3-5 actionable conclusions from this comparison.

Prompt 10: Executive Summary Generator

Use case: Produce a polished, presentation-ready summary of your review data that you can share with stakeholders, include in reports, or use for decision-making meetings. This prompt is designed to run after you have already explored the data with other prompts.

You are a senior business analyst preparing a customer feedback report. I have uploaded [NUMBER] reviews from [PLATFORM] for [PRODUCT/COMPANY NAME], spanning [DATE RANGE].

Produce a comprehensive executive summary formatted for presentation to leadership:

EXECUTIVE OVERVIEW (2-3 paragraphs)
A high-level narrative of what customers think about this product/service. Include the overall sentiment, the most significant positive and negative themes, and any notable trends.

KEY METRICS
Present in a clean table format:
- Total reviews analyzed
- Average star rating
- Sentiment split (% positive / neutral / negative)
- Net Promoter indicators (% who explicitly recommend vs. warn against)
- Top 3 strengths (with % of reviews mentioning each)
- Top 3 weaknesses (with % of reviews mentioning each)

CRITICAL FINDINGS (bullet points)
List the 5 most important findings that leadership needs to know. Each should be one sentence with a supporting data point.

COMPETITIVE POSITION
If reviewers mention competitors, summarize how this product is positioned relative to alternatives. If no competitors are mentioned, note that.

RISK FACTORS
Identify 3 risks suggested by the review data (e.g., emerging quality issues, competitor threats, declining satisfaction trends).

RECOMMENDED ACTIONS
List 5 prioritized actions with expected customer impact for each.

APPENDIX: TOP QUOTES
Include the 5 most positive and 5 most negative verbatim quotes that best represent overall customer sentiment.

Format everything with clear headers, bullet points, and tables. This should be ready to paste into a slide deck or report document.

Claude vs. ChatGPT for Review Analysis

Both Claude and ChatGPT are capable review analysis tools. If you have read our ChatGPT review analysis guide, you already know ChatGPT's strengths. Here is an honest comparison to help you choose the right tool -- or use both strategically.

When to use Claude: You have a large dataset (1,000+ reviews), you need precise categorization, you want structured output that follows your exact format, or you are doing ongoing analysis with the Projects feature.

When to use ChatGPT: You need quick visualizations through Code Interpreter, you are working with smaller datasets, or you prefer a more conversational analysis style.

When to use both: For high-stakes analyses -- competitive positioning, product launch decisions, major strategy pivots -- running the same prompt through both Claude and ChatGPT and comparing outputs produces the strongest results. Areas where both tools agree are high-confidence findings. Areas where they diverge are worth investigating further.

Advanced Techniques

Once you are comfortable with the basic export-upload-prompt workflow, these advanced techniques will deepen your analysis.

Projects for Ongoing Monitoring

Instead of running one-off analyses, set up a Claude Project for continuous review monitoring. Pin your latest review export to the project. Each week or month, export fresh reviews with Comment Exporter, update the pinned file, and run your standard prompts. Over time, you build a longitudinal view of customer sentiment that tracks the impact of product changes, marketing campaigns, and competitive moves.

Create separate projects for different analysis contexts: "Product A Customer Feedback," "Competitor Tracking," "Content Ideas from Reviews." Each project maintains its own pinned files and instructions, keeping your analyses organized and persistent.

Combining Reviews from Multiple Platforms

The most powerful insights come from cross-platform analysis. Export reviews from Amazon, comments from Reddit, and feedback from Etsy for the same product or brand. Upload all three CSV files to Claude in a single conversation.

Cross-platform analysis reveals which findings are universally true (a complaint appearing on Amazon, Reddit, and Etsy is a confirmed issue) versus platform-specific (Reddit users may surface technical issues that Amazon reviewers never mention). Claude's large context window makes this feasible -- three files of 1,500 reviews each still fits comfortably within the 200K token limit.

"I exported 2,000 Reddit comments and loaded them straight into my research pipeline." -- Pendo Kessam

Using Artifacts for Visual Reports

Claude's Artifacts feature generates structured visual elements inline. When you include instructions like "present this as a table" or "create a comparison chart," Claude produces formatted artifacts that you can:

• ✓Copy directly into Google Docs, Notion, or presentation tools.
• ✓Export as standalone documents.
• ✓Reference in follow-up conversations within the same project.

For review analysis, ask Claude to generate artifacts for sentiment breakdown tables, competitive comparison matrices, trend charts, and executive summary documents. These artifacts make it easy to share findings with team members who do not have Claude access.

Iterative Deep-Dive Analysis

Claude's conversation memory and large context window enable a powerful iterative workflow. Start broad, then drill deep:

1. ✓Run the Sentiment Overview prompt (Prompt 1) to get the landscape.
2. ✓Identify the most interesting finding -- say, 28% of reviews mention shipping problems.
3. ✓Follow up: "Show me all reviews that mention shipping issues. Group them by specific complaint type."
4. ✓Drill deeper: "For the 'delayed delivery' complaints, is there a pattern by month? Did these spike after a specific date?"
5. ✓Extract action items: "Based on the shipping analysis, write a brief for the operations team with specific recommendations."

Each follow-up builds on the previous analysis. Claude maintains full context of both the original data and all prior conversation turns, so you never lose information as you drill deeper. This iterative approach often surfaces insights that a single prompt would miss.

Tips for Better Results with Claude

These tips come from extensive testing of review analysis workflows in Claude. Apply them to get consistently better output from every prompt.

• ✓Be specific about output format. Claude follows formatting instructions precisely. If you want a numbered list, say so. If you want a table with specific columns, define them. If you want bullet points under each heading, specify that. The more explicit your format requirements, the more usable Claude's output will be.
• ✓Provide context about your product and industry. Add a sentence like "This is a mid-range Bluetooth speaker competing with JBL and Bose" or "We are a B2B SaaS tool for project management." Context helps Claude interpret ambiguous review language correctly and produce more relevant categorizations.
• ✓Ask for structured output with headers. Prompts that request structured sections (numbered lists, named headers, tables) produce more actionable output than open-ended prompts. Compare "analyze these reviews" with the structured prompts in this guide -- the difference in output quality is dramatic.
• ✓Use Claude's follow-up capability. Do not try to extract everything in one prompt. Run a broad analysis first, then ask follow-up questions to drill into specific findings. Claude maintains the full conversation context, so follow-ups are seamless.
• ✓Request direct quotes. Always include "provide verbatim quotes from the reviews" in your prompts. Quotes ground the analysis in actual data and prevent Claude from generalizing or inventing patterns.
• ✓Pre-filter for sharper insights. Analyzing 1-2 star reviews separately from 4-5 star reviews produces sharper findings than dumping everything together. Use Excel or Google Sheets to filter before uploading, or ask Claude to segment the data: "Analyze only the reviews with 1-2 star ratings."
• ✓Leverage the Projects feature for repeated analysis. If you analyze reviews regularly -- weekly competitor monitoring, monthly customer feedback reviews -- set up a Claude Project. Pin your standing instructions and update the data file as needed. This saves time and ensures consistency across analyses.

Frequently Asked Questions

How many reviews can Claude analyze at once?

Claude's 200K token context window can handle approximately 5,000 or more reviews in a single upload, depending on average review length. Short reviews (1-2 sentences) allow for higher counts; longer, detailed reviews reduce the maximum. This is significantly more than most other AI tools, which start struggling above 1,000 reviews. For datasets exceeding 5,000 reviews, you can split them into two uploads or pre-filter in Google Sheets to focus on the most relevant subset. Claude Pro users can also use Projects to maintain context across multiple conversations with the same data.

Do I need Claude Pro to analyze review files?

No. Claude's free tier supports file uploads and provides access to Claude 3.5 Sonnet, which is fully capable of review analysis. The limitation is usage -- free-tier users have a daily cap on messages, which may restrict how many analysis prompts you can run in a single day. Claude Pro ($20/month) gives you significantly higher usage limits, access to Claude Opus 4 for more nuanced analysis, and the Projects feature for pinning data files across conversations. If you plan to analyze reviews regularly, Pro is worth the investment.

What file formats can I upload to Claude for review analysis?

Claude supports CSV, PDF, and plain text file uploads. For review analysis, CSV is the recommended format because it preserves structured columns -- review text, star ratings, dates, verification status, and other metadata -- that Claude can parse and reference throughout the analysis. Comment Exporter exports directly to CSV, making the workflow seamless. You can also paste smaller datasets (under 100 reviews) directly into the chat window without uploading a file.