- Use this AI prompt to slash hours from your review-mining process
- For conversion copywriters, customer research experts, product marketers
- Analyze reviews in the time it takes you to drink your morning coffee (unless you’re chugging it)
- Perfect for ChatGPT 4 – yes, you can use it for review mining
Review mining is a great way to get voice-of-customer data. But documenting the insights can take precious hours—even days.
As you pore over hundreds of reviews on G2 and Capterra, you painstakingly copy-and-paste the reviews into a Google Sheet. You create all kinds of buckets as you go—desires, benefits, pain points. You slooowwwly identify patterns. You start sorting and ranking.
You decide you’d rather make a living fostering kittens.
Well, you can rescue your kittens and do your research too. In minutes.
Here’s what to do: Prompt AI to turn negative reviews into a detailed, organized list of pain points. This is what ChatGPT 4 created for me when I used the AI prompts I’m about to teach you:
Here’s how to use ChatGPT 4 for review mining. And get similar results for your business or client.
Step 1: Select reviews based on your research objective
Go to your review site(s) of choice. Decide what to analyze and why. Your research should always have a reason. (Remember: garbage in, garbage out.)
Decide what filters you want to apply to the reviews. For example, you may want to concentrate on 3-star reviews because they’re the juiciest. Or you may want to exclude reviews more than two years old. Or you may want to limit reviews to users within a certain industry or role.
Bottom line, it’s important to define your review-mining objective. Include or exclude reviews accordingly.
For my example, I decided to analyze reviews for a specific invoicing software product: Dubsado. I included reviews no older than 2020. And since I’m analyzing pain points, I kept it to just the “Cons” and “Dislikes” of each review.
Step 2: Create a master document of the reviews you want ChatGPT to mine for you
Use your preferred way to copy the reviews to a separate document like a Google Doc, Word doc, or other word processor.
I like Airstory for this. The Airstory-clipper Google Extension cuts down the time it takes to copy, paste, and toggle. Plus, it retains info about where each review came from.
Regardless of how you do it, make sure you have a .docx or PDF file ready to go before you proceed to the next step. And don’t worry about the length of the doc, because here’s what we’re going to do next.
Step 3: Feed that sh*t into ChatGPT
My Airstory document showed me I’d copied a total of 86 reviews. I exported it as a .docx file. The doc ended up being over 4,000 words and over 20,000 characters!
ChatGPT has a character limit per prompt. At the time of writing, the known limit is 4,096 characters per prompt.
So you’ve got a couple of options. You can either:
- Do your own manual “chunking” (yup, tedious!)
- Use a free ChatGPT chunking tool!
I used a free chunking tool called ChatGPT Splitter. And let me tell you, it worked marvelously.
- Upload your master .docx or PDF into ChatGPT Splitter
- Set your “chunk size” (e.g., 4,000 characters)
- Click the “Process” button
And voilà! The splitter creates the exact sequence of prompts you need to paste into ChatGPT. No need to create your own chunks or prompts.
As you can see, in my case, the splitter created 7 copy-and-paste prompts. Each has a “Copy” button below it. It even turns the prompts green after you’ve copied them, so you don’t accidentally skip or duplicate anything.
Important: Copy Prompt #1 into ChatGPT. Enter it. Your bot will give you an “OK” message. Repeat this for each prompt down the line. ChatGPT will wait for you to enter them one by one, like the good bot it is. (“Aaaaw, how cute!” But not nearly as cute as foster kittens.)
Once you’ve pasted and processed each prompt in ChatGPT, you should see something like this in your ChatGPT thread:
That’s ChatGPT confirming it has properly ingested the final chunk (in my case, 7 out of 7) and is ready for your next instructions. Yay.
Step 4: Ask ChatGPT to analyze the reviews
Now it’s time for analysis without the paralysis.
Copy-paste the following prompt into ChatGPT, being mindful to fill in the blank where it says “[X]”:
The above text contains [X]. Create an aggregated list of complaints or pain points. Sort the list in order of most common complaint to least common complaint. Be thorough and detailed.
For “X” I like to set the context for ChatGPT because it gives me more focused outputs. For my example today, I told ChatGPT that it has just ingested customer reviews for invoicing software:
Step 5: Pat yourself on the back. The day is young and you’ve still got a few sips of coffee left!
You’ve just used ChatGPT for review mining and distilled 86 reviews of a product before you even finished your morning latte. And you can easily process a lot more than 86 at a time. You’ve effectively cut down hours of review analysis.
You already have your buckets/categories of, in this case, pain points. Now you can go ahead and use the insights to whatever it is you were going to do next. Like turning “dozens of bad reviews into hundreds of new customers.”
Use cases to use ChatGPT for review mining
There are so many ways you can use this prompt.
Here are some ideas:
- Analyze a competitor’s product. Why? So you know what pain points to hit when you write copy that helps customers switch to your product.
- Analyze your client’s (or employer’s) product. Why? Maybe your client told you they’ve fixed a certain bug or released a new feature. Now you can use this data to write copy that overcomes objections.
- Analyze products across an entire category. Why? Maybe you’re writing copy for a disruptive product that is way better than current category leaders. You’ll want to know all the important pain points to hit when you position your client’s product as superior.
I tried this AI prompt on a more technical product. Here’s what happened.
I got ChatGPT to analyze 92 pieces of negative feedback about Boomi, an integration-platform-as-a-service (iPaaS), a middleware product. The user feedback, as you can imagine, was much more technical than for invoicing software. And yet, following these prompts, ChatGPT produced a comprehensive analysis that’s easy to understand:
From start to finish, the entire exercise took me 28 minutes, and the bulk of that time was spent on clipping/copying.
I have used this method to analyze other products as well, and each time I got a rich result—as long as I was feeding ChatGPT the information right in the chat. Here’s what I mean:
Tips and FYIs if you want to use ChatGPT for review mining
PDF plugins for ChatGPT:
There’s another way you can feed long text into ChatGPT that doesn’t involve breaking the review text into chunks or feeding them straight into the chat. There’s a pdf reader plugin, AskYourPDF. This plugin allows you to upload a PDF, and then it generates an “ID” of your doc. You then paste the document ID into ChatGPT.
Yes, it’s fast and easy. But I tried it a few times and was disappointed with the results. ChatGPT’s analysis was much thinner using this method. It missed several pain points, and the descriptions lacked detail.
Word of caution on privacy and intellectual property
Remember, as a marketer or conversion copywriter or researcher, never use AI in any way that violates contracts, agreements like NDAs, or just good old ethics.
Make sure that whatever you’re pasting into ChatGPT doesn’t violate the privacy laws of your employer or client. For example, do not paste company secrets or intellectual property into ChatGPT.
ABR: Always be responsible. Keep that brain switched on.