Competitor reviews are one of the best free sources of product ideas. Buyers tell you, in their own words, what broke, what was missing and what they wish they had bought instead. The problem is volume. Reading hundreds of reviews across five listings takes hours, and it is easy to remember the dramatic ones and miss the patterns.
AI assistants such as ChatGPT and Claude are good at this kind of work. They can group many comments into themes, count how often each theme appears and pull out example quotes. This guide shows a step-by-step method with prompts you can copy, and explains how to check that a gap is worth acting on.
What review analysis can and cannot tell you
| It can tell you | It cannot tell you |
|---|---|
| Common complaints, in buyers' own words | How many units a product sells |
| Features buyers value and expect | Whether demand is growing or falling |
| Uses the seller did not plan for | Your margin after fixing the problem |
| Questions the listing does not answer | Whether reviewers represent all buyers |
Reviews show problems. Data on demand, competition and margin shows whether solving them is a business.
Step 1: Choose the right listings
Pick 3 to 5 competing products that are actually selling. A listing with a strong, steady sales rank tells you about real buyers. One with few sales tells you little. Our guide to Amazon Best Seller Rank explains how to read it.
Include a mix: the market leader, one or two mid-priced options and a budget option. Gaps often sit between price levels.
Step 2: Collect review text you are allowed to use
Be careful here. Read the marketplace's terms before you collect anything, and follow them. In practice:
- Read reviews on the product page and copy the ones relevant to your research by hand, for your own analysis.
- Don't use scrapers or bots that break the marketplace's terms.
- Don't republish reviews. Remove reviewer names before you paste text into an AI tool.
- If you use a third-party tool that exports reviews, check that it has the right to provide that data.
Focus on 1 to 3 star reviews for problems and 4 star reviews for "good, but" comments. The 4 star reviews are often the most useful: the buyer liked the product but tells you exactly what would make it better.
Collect enough that patterns start to repeat. A few dozen reviews per listing is a sensible first pass. Add more if the themes are still unclear.
Step 3: Format the text
AI tools work better with tidy input. Put one review per line, with a listing label and the star rating. These two lines are illustrative:
A, 2 stars: Lid cracked after a month of daily use.
B, 4 stars: Works well, but the seal is hard to clean.
Label each listing with a letter instead of the brand name, and keep a note of which letter is which. If you have a lot of text, split it into batches so nothing gets cut off.
Step 4: Run the analysis prompts
Use one chat per product idea and run these prompts in order.
Prompt 1: Find the themes
Below are customer reviews for three competing products, labelled A, B and C. Group the complaints into themes. For each theme, give a short name, the number of reviews that mention it, which products it affects, and two short quotes as evidence. Only use what is in the text. Do not add themes from general knowledge.
[paste reviews]
Prompt 2: Separate product problems from everything else
Sort the themes into four groups: product design or quality, packaging and delivery, listing accuracy (the product was not as described), and seller service. Which themes could a better product fix?
Delivery and service complaints matter, but they are not product gaps.
Prompt 3: Find what buyers love
Now list what buyers praise most, with quotes. Which features would a new version need to keep to avoid new complaints?
Prompt 4: Find unplanned uses
Which reviews mention using the product for something other than its main purpose, or by a different type of buyer than the listing targets? List each use with a quote.
Unplanned uses can point to a new angle, a variant or a better listing.
Prompt 5: Turn themes into product changes
For the top 5 product themes, suggest a specific change that would solve each one. For each change, say whether it is likely to raise the unit cost, need new moulds or tooling, or affect certification. Do not estimate exact costs.
Prompt 6: Check the counts
Recount the reviews for each theme and list the line of every review you counted. Flag any review you placed in more than one theme.
AI counts can be off, especially on long inputs. This prompt makes the assistant show its work so you can spot-check it.
Step 5: Score the gaps
Put the results in a simple table. Here is an illustrative example for a reusable food container:
| Theme | Listings affected | Fixable by design? | Likely cost impact | Already solved by a competitor? |
|---|---|---|---|---|
| Lid seal leaks | A, B, C | Yes, better seal | Small | No |
| Stains from tomato sauce | A, C | Partly, different material | Medium | Yes, glass versions |
| Hard to stack | B | Yes, new shape | New tooling | No |
| Arrived damaged | B | No, packaging issue | Small | Not relevant |
The strongest gaps appear across several listings, can be fixed with a product change, don't need costly new tooling and aren't already solved by a strong competitor. In this example, the leaking seal would be the first to investigate.
Step 6: Check the gap against real data
A real complaint is not enough. Before you contact a supplier, check:
- Demand. Is the category selling steadily, or are you fixing a product few people buy?
- Competition. If one listing already solved the problem and dominates, the gap may be closed.
- Margin. A better seal or thicker material changes your unit cost. Run the full numbers, including shipping, duty and fees. Our guide to landed cost calculation shows how.
- Supplier ability. Ask suppliers whether they can make the change and what it costs. Our guide on finding and vetting a supplier covers what to ask.
Then work through the full validation checklist before you order stock.
Common mistakes
- Trusting the counts blindly. Ask for quotes and line references, then spot-check.
- Too few reviews. A theme mentioned twice may be noise.
- Forgetting who writes reviews. Very unhappy and very happy buyers are more likely to leave one. Reviews show what matters to buyers, not exact rates.
- Letting the AI fill gaps. Tell it to use only the text you provided.
- Fixing the wrong problem. A delivery complaint does not need a new product.
- Copying a competitor's wording. Use reviews to understand problems, then describe your solution in your own words.
Where Ecomtoro fits
Ecomtoro does not analyze reviews. Use ChatGPT or Claude for that step, with the prompts above.
Ecomtoro covers the step after: checking whether the gap sits in a market worth entering. It researches products using Amazon sales-rank and price history, Google Trends, Google Shopping and search volume. It builds validation dossiers with demand, trend, competition, margin and risk scores, each with a source. It also calculates profit for Amazon FBA, Shopify and TikTok Shop in the US, UK and Germany, including VAT, duty and fees.
You can use it in the app or connect it to Claude or ChatGPT over MCP, so the review themes and the market numbers sit in the same conversation.
Frequently asked questions
Is it allowed to analyze Amazon reviews with AI?
It depends on how you collect the text. Read and follow the marketplace's terms, avoid tools that break them, remove reviewer names and don't republish reviews. Summarising themes for your own research is the safest approach.
Which AI is best for analyzing reviews?
ChatGPT and Claude both handle it well when you give them clean text and clear instructions. The prompt matters more than the tool: tell it to use only your text, quote evidence and show its counts.
How many reviews do I need?
Enough for themes to repeat across listings. A few dozen low and mid-rated reviews per listing is a reasonable first pass; add more if the themes are unclear.
Does Ecomtoro analyze reviews?
No. Ecomtoro focuses on demand, trend, competition, margin and risk data with sources. Use an AI assistant for review themes and Ecomtoro to check whether the market and margin support your idea.
