Build a market report with ChatGPT Feed your AI real Google Maps data

A report is only as good as its input

Most people try to build a local market report by pasting a vague prompt into an AI and hoping it invents accurate numbers. It cannot. Language models do not know how many dentists operate in a given postcode, what their average rating is, or which ones have no website. They need real data to reason over.

That is the whole idea behind feeding your AI with business data. Instead of asking a model to guess, you hand it a clean, structured file of the exact businesses you care about. The free Vonsel Chrome extension is how you get that file: search Google Maps normally, capture the results, and export them in seconds.

The extension is the means, not the end. It pulls names, addresses, phone numbers, websites, categories, ratings, and review counts straight from the map. Once that sits in a spreadsheet, ChatGPT and Claude Code can do the heavy lifting: counting, grouping, ranking, and writing the narrative around the numbers.

Add Chrome Extension

Free download. No trial, no credit card.

Upload the file, then prompt it

ChatGPT is the fastest way to get from raw export to first insight. When you attach a spreadsheet, its code interpreter runs real Python behind the scenes to read your rows, compute statistics, and draw charts. You never touch the code. You just describe what you want.

Export your captured businesses as CSV from the extension, open a fresh ChatGPT chat, attach the file, and paste a prompt like the one below. Be specific about the market and the output you expect.

ChatGPT returns a structured answer with tables and often a chart or two. From there you iterate in plain language: "turn the rating breakdown into a bar chart", "list the 10 lowest-rated gyms with their addresses", or "rewrite the summary for a pitch deck". Because the model is working from your file, the numbers stay grounded in real listings rather than guesses. For a deeper walkthrough of prompting over a list, see analyzing lead lists with ChatGPT.

One tip: keep files focused. A single city and a single category per report gives cleaner output than dumping 20,000 mixed rows into one chat. If you need to compare markets, run several exports and ask the model to merge them.

When you want a report you can regenerate

ChatGPT is great for exploring. Claude Code shines when the report needs to be repeatable. It runs in your terminal, sits next to your exported files, and can write a small script that turns any Vonsel export into the same polished report every time. Drop next month's fresh CSV in the folder, run the command, and the market report rebuilds itself.

The prompt style is similar, but you point Claude Code at the file on disk and ask it to produce artifacts you keep. For example: "Read gyms-madrid.csv, compute the market stats, and write market-report.md with tables plus a ratings chart saved as a PNG." Because it can create and run files, it is also the natural choice for bigger datasets that a chat window would struggle with.

This is where the two tools split cleanly by job. Many people explore in ChatGPT to find the angle, then hand the pattern to Claude Code to automate. If you want to go further and turn the numbers into something interactive, our guide on building a lead dashboard with Claude Code picks up right where a static report ends.

Codex and local models for the same file

ChatGPT and Claude Code cover most people, but the exported file is not tied to any one tool. OpenAI Codex works much like Claude Code in the terminal if that is your ecosystem: point it at the CSV and ask it to script the report. The workflow is identical, only the assistant changes.

If your data is sensitive or you simply prefer to keep it off the cloud, a local model can read the same file. The tradeoff is quality: a small local model handles counting and grouping well but writes a weaker narrative than a frontier model. For most market reports that is an acceptable compromise, and your rows never leave your machine.

Whichever assistant you pick, the format matters. CSV and Excel are best for flat tables and charts. JSON keeps nested fields like reviews and opening hours intact, which is handy when you want the model to reason over review text. The extension exports all three, so you can match the file to the question. Our note on JSON vs CSV for AI analysis breaks down when to reach for each.

Sections that make a report useful

From a one-off report to a working pipeline

A CSV in ChatGPT is perfect for a quick study. But once the report tells you which businesses to chase, you need somewhere to act on them, and a spreadsheet is not it. That is the end the extension is a means toward.

Inside the Vonsel dashboard, the same captured businesses become a mapped CRM. Every listing sits on the map, tagged by status. Review Intelligence reads the reviews for pain points, Sales Intelligence scores the fit, and Email Intelligence drafts a tailored email per business from what it found. You can reopen a capture later to add more, attach reviews to businesses you already have, and enrich websites for contact details, so the database gains context rather than going stale. It is the difference between a value-added database and a raw CSV.

Use the AI report to understand the market. Use the dashboard to work it. If your angle is pure research for clients, the same data feeds a reviews-driven market research deliverable you can charge for.