I have not briefed a copywriter for ad creative in months. Not because good copywriters do not exist, but because the biggest bottleneck in DTC ad performance is not writing quality. It is data quality. And no copywriter, however talented, reads 800 customer reviews before drafting a single line.
The VOC ad copy agent does. It reads every review it can find, identifies the language patterns customers repeat, maps their objections and desired outcomes, and converts all of that into direct-response ad copy and HTML creative mockups, in 22 minutes, without a single brief.
What writing DTC ad copy used to cost
A standard creative round with a DTC copywriter or agency runs between £2,000 and £5,000 and takes two to three weeks. You write a brief, someone interprets it, you review a draft, you revise, you brief the designer, the designer interprets that, and by the time the creative hits Meta it carries three layers of interpretation between the brand and the customer. And it almost always uses brand language, not customer language, because the people who wrote it were briefed by the brand, not by 800 actual buyers.
The deeper problem is that most DTC brands have the data to write better ads sitting untouched in their reviews. Customers tell you exactly what made them buy, what almost stopped them, what surprised them, and what they tell their friends. That language converts because it sounds like the customer talking to themselves, not a brand talking at them. Before the agent, extracting and using that data was a manual research project that almost nobody had the time or discipline to do consistently.
What the VOC ad copy agent actually does
The workflow runs in six steps. Here is exactly what happens from input to output.
Input: one URL or a review CSV
The agent takes a Shopify product URL, an Amazon ASIN, a Trustpilot brand page, or a raw review CSV. Most runs use a combination. For brands with no direct review presence, a Reddit or forum crawl can substitute. The input does not need to be clean or formatted. The agent processes it.
Scrape: 800+ reviews across sources
It crawls Trustpilot, Amazon, Shopify native reviews, and Reddit brand mentions in parallel. For a typical DTC brand with 12 months of review history, this pulls between 400 and 1,200 individual customer responses. The scrape is rate-limited and source-labelled so the analysis can weight by platform where needed.
Cluster: themes, objections, and exact language
The agent groups all review text by semantic theme. Not by keyword matching, by meaning. It identifies the top 6 recurring objections (the things that nearly stopped people buying), the 4 strongest desire statements (the outcomes customers celebrate), the 3 core benefit clusters, and the exact phrases customers repeat. These are the phrases that will appear verbatim in the ad copy.
Build: the VOC framework
From the clustering output, the agent builds a structured voice-of-customer brief. This includes ranked objections by frequency, desire statements by emotional weight, and a vocabulary map: what customers call the product, the problem it solves, and themselves. This document is the creative brief. It replaces the traditional briefing process entirely.
Write: 5 direct-response ad variants
Using the VOC brief and the brand context layer, the agent writes one ad variant per framework: PAS (Problem, Agitate, Solution), AIDA (Attention, Interest, Desire, Action), curiosity hook, social proof lead, and objection-flip. Every line is traceable to a real review cluster. Nothing is invented. The copy sounds like a customer because it was written from customer data.
Produce: 15 HTML creative mockups
For each of the 5 variants, the agent generates HTML creative mockups in 3:4 (Meta feed), 9:16 (Reels and TikTok), and 1:1 (Instagram square) formats. These are not rough wireframes. They are formatted, on-brand assets with headline, body, CTA, and product image placement, ready to export and push into an ad account for testing.
The brand context layer: why it does not write generic copy
Most AI copywriting tools produce generic output because they have no brand context. They know about writing frameworks. They do not know that your supplement brand avoids clinical language, that your price point means the copy must address value before it addresses ingredients, or that your primary customer archetype is a 34-year-old woman who has already tried three competitors and is sceptical of marketing claims.
The VOC agent runs against a brand context document that is built once and refined with each run. It contains the product category, the price positioning, a vocabulary map of approved and banned phrases, tone calibration (clinical, conversational, or lifestyle), the primary customer archetype, and the three outcomes that drive repeat purchase. When the agent writes, it writes within that context. The output sounds like the brand because it is calibrated to the brand, not just to the product description.
What the output actually looks like
Here is a real example from a protein supplement brand run. The agent processed 847 reviews across Amazon, Trustpilot, and a Reddit thread. These are the top clusters it surfaced.
VOC cluster output — protein supplement brand
"no bloating" or "doesn't bloat me"
Primary outcome
"actually mixes properly" or "no clumps"
Objection overcome
"doesn't taste chalky" or "tastes like actual food"
Category objection defeated
"expensive but worth it" or "worth paying more"
Price objection + worth-it frame
"I've tried everything" or "switched from X"
Switcher archetype signal
From those clusters, here is what the agent produced for two of the five variants.
Variant 01 — PAS hook
"Most protein powders leave you bloated and lying to yourself about whether they're actually working. This one doesn't. Here's why 12,000 athletes switched, and why none of them went back."
Variant 03 — Curiosity + social proof
"The reason this protein has 4,200 reviews averaging 4.8 stars isn't the ingredients list. It's what happens in your gut at the 30-day mark. The people who stuck with it know. Read what they said."
Each variant is output alongside three HTML creative mockups in 3:4, 9:16, and 1:1 formats, with the headline, hook, body copy, and CTA already placed in the correct positions for each format. The person running the test does not need to brief a designer. They export the mockup and push it.
Inside the system
How we build this for brands
The VOC agent is one component of the broader creative system we run for portfolio brands. It feeds directly into the paid media workflow: review data surfaces the angles, the agent writes and formats the creative, and the TikTok and Meta image generation tools produce the visual assets. The output goes into an ad account with a structured testing framework that tells us within 48 hours which angle is worth scaling. The cycle from raw review data to a winning creative variant now takes days, not weeks.
When we take a brand on, we also connect the VOC engine to the email system. The same customer language that builds ad hooks feeds the Klaviyo flows, so the messaging a customer sees in a Meta ad is consistent with what lands in their inbox. That consistency is what turns a first click into a second purchase. Part of this runs live for portfolio brands today; the full system is what we design and deploy when we take a brand on.
Build Your Agent Team
Get your own AI agent team working on your brand
Caner can help you build your own AI agent teams and unlock growth for your brand. This is not a template or a SaaS tool. It is a custom system designed around your data, your workflows, and your growth targets.
Book A DemoFrequently asked questions
What is voice of customer data and why does it improve ad performance?
Voice of customer data is the actual language your customers use when describing their experience with your product: the phrases they repeat in reviews, the objections they raise before buying, the outcomes they celebrate after. It improves ad performance because it eliminates the gap between what brands say about themselves and what customers actually care about. When ad copy uses the customer's own words, it triggers recognition rather than persuasion, and recognition converts at a much higher rate.
Can I build this VOC ad copy agent myself?
The components exist as standalone tools: review scrapers, AI copywriting tools, design generators. But the agent is not a single tool. It is a system built on a custom orchestration layer, a brand context document, a trained voice calibration for each brand, and a creative brief format that connects VOC themes to specific ad frameworks. The integration work and the brand configuration are what make it produce usable output rather than generic copy. You can approximate pieces of it manually, but the speed and consistency of the automated version comes from the system, not the individual components.
How long does it take to set up the VOC agent for a brand?
For brands I work with, the initial setup takes one session: gathering the product URLs, configuring the brand context document, calibrating the tone and any copy constraints, and running the first full pass. The first output is reviewed and refined. Subsequent runs are fully automated and take 22 minutes end to end. The setup is not a long onboarding process. It is a focused configuration that makes every future run consistent and brand-accurate.
What review sources does the agent pull from?
The agent pulls from Trustpilot, Amazon (for brands selling there), Shopify native reviews, Reddit brand mentions, and any review CSV you provide directly. For most DTC brands, the richest source is Amazon reviews because they tend to be longer and more specific. Shopify reviews are useful for tone calibration. Reddit surfaces the unfiltered objections that customers would not write in a formal review.
How many ad variants does it produce per run?
Each run produces 5 direct-response ad variants using different copywriting frameworks: PAS, AIDA, curiosity hook, social proof lead, and objection-flip. For each variant, it generates 3 HTML creative mockups in 3:4, 9:16, and 1:1 formats. That is 15 ready-to-test assets from a single run, all grounded in real customer language rather than brand assumptions.
Does the VOC agent replace a copywriter?
For the first-draft, volume-testing layer of ad creative, yes. The agent produces copy that is grounded in real customer language and formatted for immediate testing. What it does not replace is strategic creative direction: knowing which angle to prioritise for a new product launch, reading the cultural moment for a seasonal campaign, or writing for an emerging persona that reviews have not yet captured. The agent handles the data-to-draft workflow. The operator handles the strategy that points it in the right direction.
About the author
Caner Veli built Liquiproof to global distribution across 3,000+ retailers, then exited. He now runs Purposeful Profits using a combination of operator strategy and AI-powered systems he has built and uses daily, having 10x'd monthly revenue in his own business in the last 90 days.