AI Customer Feedback Analyst Prompt
Introduction
Picture this: you’ve just wrapped up a product launch, and the reviews are flooding in. Hundreds of them. Some are glowing, some are brutal, and a whole lot are somewhere in the middle. Now someone on the team needs to make sense of all of it — fast. Without the right system, you’re staring at a wall of text, manually reading through review after review, trying to spot patterns that may take hours to surface. It’s exhausting, and worse, it’s inconsistent. Human readers bring bias. They skim. They miss things.
This is exactly where AI changes the game. With the right prompt, you can turn a messy pile of raw customer reviews into a clean, structured breakdown — sorted into Positive, Negative, and Mixed categories, complete with summaries — in the time it takes to grab a coffee. No guesswork. No cherry-picking. Just cold, neutral analysis driven entirely by what your customers actually said.
The AI Customer Feedback Analyst prompt you’re about to get was built for precision. It follows a strict step-by-step workflow, refuses to inject assumptions, and outputs a clear quantitative summary at the top so you always know exactly what you’re working with. Whether you’re a solo founder reading Etsy reviews or a marketing manager combing through thousands of product responses, this prompt turns feedback chaos into clarity.
The Master Prompt
Copy and paste the prompt below directly into Claude AI to get started. No setup required — just paste and go.
Act as an Expert Customer Feedback Analyst. Your task is to analyze customer reviews, categorize them into "Positive", "Negative", and "Mixed", and provide concise summaries for each. You must follow this exact step-by-step workflow: Step 1: Request Input Respond with this exact message (no extra text): "Hello! I'm ready to analyze your customer feedback. Please paste the reviews as plain text below (or upload a file if your platform supports it). Let me know when you're ready." DO NOT perform any analysis yet. Stop generating text and wait for my response. Step 2: Confirmation Once I provide the data, DO NOT analyze it immediately. Instead, ask exactly: "I have received your data. Is there anything else you want to add, or should I proceed with the analysis now?" If I provide more data or state an issue, address it politely and ask for confirmation again. Wait for explicit confirmation such as ‘proceed’, ‘yes’, or ‘go ahead’ before continuing to Step 3. Step 3: Analysis and Output Analyze the provided data based strictly on these rules: * Base all analysis exclusively on the provided reviews. Do not add any external knowledge, assumptions, or interpretations beyond what is explicitly stated in the data. Remain 100% neutral and data-driven at all times. * If the input is empty or not recognizable as customer reviews, request valid data. * Categorize the feedback into exactly three groups: Positive, Negative, and Mixed. * Positive: Reviews that are clearly positive with no significant negative elements. * Negative: Reviews that are clearly negative with no significant positive elements. * Mixed: Contains both positive and negative sentiment in the same review, regardless of proportion. * For reviews with mixed sentiment, place them in the Mixed category rather than splitting them. * If there are many reviews (especially over 100), group similar themes and use only the most representative key excerpts rather than listing every review verbatim to keep the output readable and concise. * Preserve the original wording of review excerpts (do not paraphrase). * If there are no positive, negative, or mixed reviews, clearly state "No positive reviews found.", "No negative reviews found.", or "No mixed reviews found." in the relevant section. * Each review must be counted once and only once. * Ignore duplicate reviews if identical text appears multiple times. CRITICAL: Do NOT provide actionable insights, recommendations, suggestions, or solutions. Stick strictly to analysis. At the very top of the final output, include this one quantitative line: "Total Reviews Analyzed: X | Positive: Y | Negative: Z | Mixed: M" (replace X, Y, Z, M with the actual counts). Format your final response as closely as possible (immediately after the quantitative line): Positive Reviews [List the positive reviews or key positive excerpts here] Negative Reviews [List the negative reviews or key negative excerpts here] Mixed Reviews [List the mixed reviews or key mixed excerpts here] Positive Summary [Summarize recurring themes only (not individual reviews).] Negative Summary [Summarize recurring themes only (not individual reviews).] Mixed Summary [Summarize recurring themes only (not individual reviews).]
How to Use This Prompt
This prompt is built exclusively for Claude AI, available at claude.ai. Claude’s ability to follow multi-step instructions with strict formatting makes it the ideal tool here — other AI tools may skip steps or blend the workflow together, which can throw off the accuracy of the analysis. Stick with Claude for the best results.
For the highest level of precision, run this prompt in batches of 100 to 200 reviews at a time. If you’re working with a larger dataset, start a fresh New Chat for each batch of 200 reviews. This keeps Claude fully focused on the current set and prevents any data overlap between batches, which can quietly reduce accuracy when you least expect it.
Here’s how the process works, step by step:
Open a new chat at claude.ai. Paste the entire Master Prompt into the chat and hit send. Claude will immediately respond with a ready message — it won’t start analyzing yet. That’s intentional. It’s waiting for your data.
Once you see that message, copy your customer reviews as plain text and paste them directly into the chat. If your reviews are saved in a document, you can upload the file instead — Claude supports that too. After receiving your data, Claude will confirm receipt and ask whether you want to add anything else or go straight into the analysis. Simply reply with “proceed” or “yes” and it takes care of the rest automatically.
The final output will open with a clean quantitative line — Total Reviews Analyzed, Positive count, Negative count, and Mixed count — followed by sorted review excerpts and theme-based summaries for each category. It’s structured, readable, and ready to share or drop straight into a report.
Example Output
Total Reviews Analyzed: 10 | Positive: 3 | Negative: 2 | Mixed: 5
Positive Reviews
“The product quality is outstanding! It exceeded my expectations and arrived two days early. Highly recommended!”
“Fast shipping and great packaging. I will definitely buy from this store again.”
“Best customer support I’ve ever dealt with. They solved my issue in minutes!”Negative Reviews
“Extremely disappointed. The item arrived broken, and the customer service team was very rude when I asked for a refund.”
“Absolute waste of money. It doesn’t work as advertised and feels very cheaply made.”Mixed Reviews
“The app’s interface is beautiful and very easy to use, but unfortunately, it crashes every time I try to save my progress.”
“While the features are great and very helpful, the subscription price is way too high for what it offers.”
“The food was delicious and the atmosphere was great, but we had to wait for over 45 minutes to get our table even with a reservation.”
“The laptop has a great screen and keyboard, but the battery life is very poor, lasting only 2 hours.”
“I love the design, but the material feels a bit thin. Overall it’s okay for the price.”Positive Summary
Recurring themes center around product quality exceeding expectations, reliable and fast shipping with good packaging, and highly responsive customer support that resolves issues quickly.
Negative Summary
Recurring themes include products arriving damaged or not functioning as advertised, perceived poor build quality, and unprofessional or unhelpful customer service during the refund process.
Mixed Summary
Recurring themes reflect satisfaction with design, features, food quality, and usability, consistently offset by concerns about technical reliability (crashes, poor battery life), pricing relative to value, thin or low-grade materials, and service delays despite reservations.
Here’s what the final output looks like inside Claude — a clean quantitative summary line followed by neatly separated Positive, Negative, and Mixed review sections, each with a concise theme summary below.
How to Customize This Prompt
The base prompt is powerful on its own, but it’s also easy to adapt for different use cases. You’re not locked into one style of output. A few small tweaks can shift the entire focus depending on what you need. Here are three variations worth trying.
Option 1: Product-Specific Feedback Filter
If you’re analyzing reviews for a single product or service rather than a general review pool, adding a product name and context line at the top helps Claude anchor its analysis more precisely.
Act as an Expert Customer Feedback Analyst. The reviews below are specifically for [Product/Service Name]. Your task is to analyze these customer reviews, categorize them into “Positive”, “Negative”, and “Mixed”, and provide concise summaries for each. You must follow this exact step-by-step workflow: [paste the rest of the original prompt steps here, unchanged]
Option 2: Urgency and Priority Flag Add-On
Want Claude to flag reviews that mention urgency signals like refund requests, safety concerns, or strong emotional language? Add a single instruction line inside Step 3 to surface those automatically.
Act as an Expert Customer Feedback Analyst. Your task is to analyze customer reviews, categorize them into “Positive”, “Negative”, and “Mixed”, and provide concise summaries for each. In addition, within the Negative and Mixed categories, flag any reviews that contain urgency indicators such as refund requests, safety complaints, or strong emotional language with the label [URGENT] next to the excerpt. You must follow this exact step-by-step workflow: [paste the rest of the original prompt steps here, unchanged]
Option 3: Competitor Mention Tracker
If you’re in a competitive market and want to know whether customers are name-dropping rival brands in their feedback, this version adds a lightweight detection layer without disrupting the core analysis structure.
Act as an Expert Customer Feedback Analyst. Your task is to analyze customer reviews, categorize them into “Positive”, “Negative”, and “Mixed”, and provide concise summaries for each. Additionally, at the end of the output, include a separate section titled “Competitor Mentions” and list any reviews that reference a competing brand or product by name, preserving the original wording. You must follow this exact step-by-step workflow: [paste the rest of the original prompt steps here, unchanged]
Option 4: Multi-Product Batch Comparison
Running reviews across two different products or variants at the same time? This version instructs Claude to keep them separated throughout the entire analysis, making it easy to compare results side by side.
Act as an Expert Customer Feedback Analyst. I will provide reviews for two separate products labeled “Product A” and “Product B”. Analyze them independently and produce two separate full outputs — one for each product — each with its own quantitative summary line, categorized excerpts, and theme summaries. Do not combine reviews from both products. You must follow this exact step-by-step workflow for each product: [paste the rest of the original prompt steps here, unchanged]
Conclusion
Reading customer reviews one by one is a thing of the past. This AI Customer Feedback Analyst prompt gives you a structured, neutral, and repeatable system for turning raw feedback into organized insight — every single time. No bias, no missed reviews, no vague takeaways. Just clean data you can actually use.
Paste it into Claude, follow the simple two-step setup, and watch a pile of unstructured reviews transform into a report-ready analysis in minutes. Try the base version first, then experiment with the customization options to fit your specific workflow. Once you’ve run it once, you’ll wonder how you ever handled feedback any other way.
