Customer Reviews to Benefit Image Ads

Chart which benefits 4–5 star reviewers talk about most, then generate flat Meta image ads for the themes you pick.

Copy this prompt & paste into Claude Code

Build a small tool that turns 4–5 star reviews into a benefit frequency chart, then generates flat Meta image ads for the benefits you pick — using the Static Ads Lab API.

Turn a customer reviews export into a simple benefit frequency chart, then generate flat Meta-ready image ads for the benefits you pick.

Stack: TypeScript, Express, Vercel AI SDK, Google Gemini, Static Ads Lab API.

View on GitHub

What it builds

A small Express server with a vanilla web UI. Upload a CSV of reviews, the server keeps the 4-star and 5-star rows, and Gemini extracts short signals and verbatim snippets from each review before grouping them into benefit themes with counts. You see a bar chart of those counts, choose which benefits and design templates to use, and the server generates one flat image ad per benefit × template pair.

Key patterns

Two-pass Gemini analysis — per-review signal and snippet extraction, followed by a cross-review pass that groups signals into benefit themes and counts supporting reviews
One prompt per ad combining the benefit theme and the best supporting review, so Static Ads Lab writes on-template copy without manual text-node injection
Batch flat image ad creation with parallel POST /v1/image-ads calls and batched status polling via GET /v1/image-ads?ids=...
Server-only API keys — the browser never touches Gemini or Static Ads Lab secrets
Workspace gate — the dev server shows a strip listing any missing keys or recipe constants so it's obvious what to fill in before you can generate

Prerequisites

Node.js 20+
A Google AI Studio key for Gemini
A Static Ads Lab API key with wallet balance
Brand, product, audience, and design template IDs already set up in your workspace
A CSV of customer reviews with rating and review columns

Quick start

cd recipes/reviews-to-benefit-image-ads
npm install
cp .env.example .env
# Add keys to .env and fill ids in src/config/recipe-constants.ts
npm start

Open http://localhost:9001 — upload a CSV, review the benefit chart, pick benefits and design templates, and generate ads.

For AI coding agents

Each recipe includes a prompt.md file designed to be handed directly to Claude Code, Cursor, or your preferred AI coding assistant. It contains the project's intent, architecture, edit points, and suggested next tasks.