Voice-of-Customer Research for Better Ads
Your best ad copy is already written by your customers. Here is how to collect their words from reviews, surveys and support, and turn them into ads that sell.
The best-performing line in an ad is often something a customer said first. "I stopped waking up sweaty." "It finally fits my wide feet." "My dog has not destroyed it yet." No copywriter would invent these. They are too plain, too specific, too real. That is exactly why they work.
Voice-of-customer research, or VOC, is the practice of collecting how customers describe their problems, their hesitations and the results they got, in their own words, and using that language in your marketing. It is the cheapest creative research you can do, and most brands skip it.
This guide covers where to find customer language, how to organize it, and how to turn it into ads for Meta and ChatGPT.
What you are looking for
You are listening for five kinds of statements. Each maps to a part of an ad.
| What customers say | Example | Where it goes in the ad |
|---|---|---|
| The problem before buying | "Every pillow went flat in a month" | Hook |
| What they tried before | "I tried memory foam and it was too hot" | Comparison or failed-fix angle |
| What almost stopped them | "I was worried it would be too firm" | Objection handling, offer |
| The moment it clicked | "The 100-night trial made it easy to try" | Offer emphasis |
| The result | "I sleep through the night now" | Promise, testimonial |
You want the exact phrasing. "Too hot" and "sleeps hot" and "wake up sweaty" are slightly different, and the version your customers use most often is usually the one that resonates.
Eight places to find customer language
1. Your own product reviews
Start here. Export every review you have. Focus on three-star and four-star reviews as much as five-star ones; they often contain the clearest descriptions of hesitations and trade-offs.
2. Competitors' reviews
Reviews of competing products on retailer sites and marketplaces show what buyers in your category care about, and what disappoints them. One-star and two-star reviews of competitors are a gold mine for angles, as long as you never name those competitors in your ads.
3. Post-purchase surveys
Add one or two open-ended questions to your order confirmation flow:
- "What almost stopped you from buying today?"
- "What were you using before this?"
Keep it short. One good question beats ten mediocre ones.
4. Support tickets and chat logs
Pre-purchase questions show objections. Post-purchase questions show confusion and unmet expectations. Ask your support team for the ten questions they answer most.
5. Customer interviews
Five 20-minute calls with recent buyers will teach you more than most dashboards. Ask about the moment they decided to look for a solution, what else they considered, and what nearly made them leave.
6. Forums and communities
Reddit threads, Facebook groups and niche forums show how people describe the problem when they are not talking to a brand. Search for your problem, not your product.
7. Questions people ask AI assistants
Buyers increasingly ask ChatGPT for recommendations. You cannot see other people's conversations, but you can map the likely questions: ask yourself how a buyer would phrase their problem to an assistant, then test those phrasings. The buyer question finder generates likely questions from your product and niche. Our guide to finding the questions your buyers ask ChatGPT covers this in more detail.
8. Long-running ads in your niche
Ads that have run 60 days or more in the Meta Ad Library are not customer language directly, but they show which customer language other brands have found worth paying for. Cross-reference them with your reviews.
How to organize what you collect
Raw quotes are not useful until they are sorted. A simple spreadsheet is enough.
Step 1: Paste each quote into its own row
Include the source (review, survey, ticket, interview) and the date.
Step 2: Tag each quote
Use the five categories from the first table: problem, prior solution, objection, trigger, result. A quote can have more than one tag.
Step 3: Add a theme
Group similar quotes under short themes like "sleeps hot," "neck pain," "goes flat," "too firm worry," "trial reassured." Keep theme names in customer language, not internal jargon.
Step 4: Count
Count how many quotes fall under each theme. With 100 to 200 quotes, clear patterns usually appear. Your top three or four themes by count are your strongest candidate messages.
Step 5: Pull the best lines
For each top theme, highlight the two or three most vivid, specific quotes. These become hooks, headlines and testimonial candidates.
A finished sheet might look like this for a hypothetical pillow brand:
| Theme | Count | Best quote | Type |
|---|---|---|---|
| Sleeps hot | 41 | "I stopped flipping my pillow to the cold side" | Result |
| Goes flat | 33 | "Every pillow I owned was a pancake by spring" | Problem |
| Firmness worry | 22 | "I was scared it would feel like a brick" | Objection |
| Trial reassured | 18 | "The trial is the only reason I tried it" | Trigger |
| Neck pain | 15 | "I do not wake up with a stiff neck" | Result |
Turning research into ads
Hooks from problem quotes
The problem quote often works almost verbatim as a hook, lightly edited.
- Customer: "Every pillow I owned was a pancake by spring."
- Hook: "Tired of pillows that go flat by spring?"
Promises from result quotes
- Customer: "I stopped flipping my pillow to the cold side."
- Promise: "A pillow that stays cool, so you stop flipping it."
Offers from objection and trigger quotes
If 22 people worried about firmness and 18 mentioned the trial as the reason they bought, the trial should be front and center in the offer. You might also add an adjustable fill if your product has one.
Testimonials from the best lines
With permission, the strongest result quotes become social proof. Keep them genuine and typical, as covered in our guide to using social proof without breaking rules.
Examples by channel
Meta primary text
Flipping your pillow to the cold side at 2 a.m.? Our pillow uses a breathable cover and shredded fill you can adjust. Try it for 100 nights. If it is not right, send it back.
ChatGPT chat card
ChatGPT ads appear below an answer when the conversation matches your context hints. Customer language is useful twice here: in the copy, and in the context hints themselves, which describe situations in plain language.
- Headline (33 chars): A pillow for people who sleep hot
- Body (91 chars): Breathable cover and adjustable fill so it stays cool and keeps its shape. 100-night trial.
- Context hint: "Person who wakes up hot at night and is looking for a pillow that stays cool"
The context hints generator can draft hints from your customer themes.
Common mistakes
- Only reading five-star reviews. They are flattering but thin. The detail is in the middle.
- Paraphrasing into marketing speak. "Superior thermoregulation" loses everything "stops me sleeping hot" had.
- Over-weighting one loud quote. Count themes. One vivid quote is a line; forty similar ones are a message.
- Using sensitive attributes. Do not write ads that imply you know a person's health condition, finances or insecurities. Meta restricts personal-attribute targeting language, and ChatGPT ad policies prohibit exploiting insecurity or implying the ad knows the user's conversation.
- Doing it once. Repeat VOC research quarterly. Language shifts, and new objections appear as you reach new buyers.
A 90-minute VOC sprint
If you only have an afternoon:
- Export your last 200 reviews (15 minutes).
- Skim 50 competitor reviews, mostly two and three stars (20 minutes).
- Ask support for their top ten questions (5 minutes).
- Tag and theme everything in a spreadsheet (35 minutes).
- Write ten hooks from the top three themes (15 minutes).
That gives you a research-backed starting point for your next round of creative. Pair it with the promise, proof, offer framework to turn themes into full ads.
How SecondWin uses customer language
SecondWin does part of this work automatically for ChatGPT ads. It reads your site to understand your product and buyer, studies the longest-running Meta ads in your niche to find the messages that keep selling, and maps those messages to the questions buyers ask ChatGPT. The ads it writes are original, use the proof on your own site, and are checked against OpenAI's policies before launch. Your own VOC research makes the inputs even better.
Get a free analysis of your site and niche to see which buyer questions and messages it finds. Plans start at $99 a month.
FAQ
What is voice-of-customer research?
Voice-of-customer research is the process of collecting how customers describe their problems, hesitations and results in their own words, then using that language in marketing. Sources include reviews, surveys, support tickets, interviews and online communities. The goal is to write ads that sound like the buyer's own thinking, which tends to be more specific and more persuasive than language invented by a marketing team.
How many reviews do I need for VOC research?
You can learn a lot from 100 to 200 quotes. That is usually enough for clear themes to emerge when you tag and count them. If you have fewer reviews, add competitor reviews, support questions and a few customer interviews. Five well-run interviews often surface the main objections and triggers even when review volume is low.
Can I use competitors' reviews in my ads?
You can read them for research, but do not quote them in your ads or name the competitor. Use them to understand what buyers in your category care about and what disappoints them, then address those points with your own product, proof and offer. Your ads should only quote your own customers, with their permission.
How does VOC research help ChatGPT ads?
It helps in two places. First, customer language makes the short headline and body more specific and relatable. Second, it improves context hints, which describe the situations an ad belongs in using plain language. Hints written the way customers describe their problem tend to be more accurate than keyword-style hints.