Get contact data for any business then feed it to your AI

Contact data is the raw material, not the finish line

Every AI workflow that touches local businesses runs into the same wall first: you need real contact data. A model can score leads, write emails, and spot patterns, but only if you hand it a clean list of businesses with phones, websites, and addresses attached. Prompting an AI to "find me dentists in Miami" gives you guesses, not a database you can act on.

So the practical move is to separate the two jobs. First capture the data from a source that actually has it, which is Google Maps. Then feed that structured list to whichever AI tool fits the task. The scraper is the means, your AI is the analyst, and the outcome is a working list.

The Vonsel Chrome extension handles the first job. You search a business type and location on Google Maps, run the capture, and it pulls names, phone numbers, websites, addresses, categories, ratings, and more. Then it exports the batch as CSV, Excel, or JSON so any model can read it. This is the literal meaning of feeding your AI with data.

Capture the list and pick your export

Open Google Maps, type something like "law firms in Austin" or "coffee shops in Lisbon," and let the results load. Click the Vonsel pop-up and start the capture. The extension walks the listings and builds a structured record for each business, then lets you export in CSV, XLSX, or JSON.

Format choice matters more than people expect once AI enters the picture. Each one has a natural home, so match the file to the tool you plan to use next.

If you want a deeper look at the tradeoffs, our note on JSON vs CSV for AI analysis covers when each one wins. The short version: CSV for conversational analysis, JSON for anything programmatic.

Feed the file to ChatGPT

For most people, ChatGPT is the fastest way to turn a raw list into something useful. Export your capture as CSV, open a chat, attach the file, and describe what you want. The model reads every row and works across the whole list, not just the part you can see on screen.

The trick is to prompt with intent. A vague "analyze this" wastes the upload. Instead, tell it exactly what job to do against your columns. Here are prompts that map cleanly to a business contact export.

From here you can iterate: ask it to analyze the lead list further, or jump straight to writing cold emails from the scraped data. Because the file already contains phones and websites, the model has real anchors to personalize around instead of inventing details.

Point Claude Code at the JSON

When your list gets bigger or you want repeatable output, a chat window starts to feel cramped. This is where Claude Code shines. It is an agentic coding tool that reads files on your machine, writes scripts, and runs them, so it can process a JSON export of thousands of businesses without you copying anything into a prompt box.

Export the capture as JSON, drop it in a folder, and open Claude Code there. Then describe the outcome in plain language and let it write and run the code.

The advantage is that the work becomes reusable. Once Claude Code writes a scoring script, you run it on every future capture. For a full walkthrough, see analyzing leads with Claude Code.

Codex, local models, and everything else

ChatGPT and Claude Code cover most needs, but the same JSON or CSV export works anywhere. If you live in a terminal, Codex can enrich lead lists with the same script-and-run approach. If privacy matters or you process large volumes on a budget, you can run local LLMs against the lead data so nothing leaves your machine.

Because the extension follows the standard Chrome extension model and exports plain files, there is no lock-in on the AI side. Whatever model you prefer today, and whatever you switch to next year, the input is a clean file of businesses. That is the whole idea behind feeding your AI with business data.

From a raw file to a working pipeline

AI analysis is powerful, but a scored CSV is still just a file. At some point you need to call the leads, track who replied, and see them on a map. That is the job the Vonsel dashboard does, and it is why the extension exists in the first place.

Inside the dashboard, captured businesses land in a mapped CRM. Review Intelligence reads the Google reviews and surfaces pain points, so you know a restaurant is getting complaints about wait times before you ever pick up the phone. Email Intelligence drafts an AI email per business grounded in those specifics. The result is a value added database, not a raw dump: contextualized with the exact angle that will land.

You can reopen a previous capture and add more businesses, attach reviews to leads you already saved, and keep everything in a searchable history. If you want the honest boundaries of what is fair game, read whether it is legal to scrape Google Maps and why Google Maps data beats bought lists.