Build a custom GPT with your own Google Maps lead database

Feed your AI with real business data

Ask a chatbot to list plumbers in your city and you get vague, half invented results. Language models are trained on the open web, not on a fresh, structured list of the businesses you actually want to sell to. The model is the engine. Your data is the fuel.

That is the whole point of this workflow. You do not need a bigger model. You need to hand the model a clean database of real businesses: names, addresses, phone numbers, websites, categories, ratings, and reviews. Once the AI can read that, it stops guessing and starts working with facts.

The fastest way to build that database is a browser capture. The Vonsel Chrome extension reads the businesses on a Google Maps search and lets you export them as a file. The extension is the means. Your custom GPT or Claude Code workflow is the end.

Capture and export your lead database

Open Google Maps and search for what you want, for example "coffee shops in Austin" or "roofers near me". Run the Vonsel extension over the results and it captures every listing on the map into a structured table. You can reopen a capture later and add more, so the database grows as you work.

When you are ready, export. The extension writes three formats from the same capture, and the one you pick depends on how you plan to use the AI. For a deeper look at the choice, see our guide on exporting leads as CSV, Excel, and JSON for AI.

If you have never done a capture, the step by step in export Google Maps data to CSV walks through it. From here on, assume you have a file with a few hundred rows sitting on your desktop.

Build a custom GPT in ChatGPT

A custom GPT is a version of ChatGPT you configure once and reuse. You give it instructions and attach files as knowledge, and it answers every question against those files. This is the no code path, and it is where most people should start.

Inside ChatGPT, open the GPT builder, give your GPT a name like "Local Lead Analyst", and upload your JSON or CSV export as knowledge. In the instructions box, tell it what it is: an assistant that answers questions about a list of local businesses, scores them for a specific offer, and drafts short outreach. OpenAI documents the builder flow and file handling in their assistants and knowledge documentation.

Now you can ask plain questions and get grounded answers:

If you would rather not build a reusable GPT and just want answers now, you can drop the file straight into a normal chat. Our walkthrough on uploading a lead CSV to ChatGPT covers that quicker version. The prompting principles are the same: be specific about the offer, and the answers stop coming back generic.

Turn it into a pipeline with Claude Code

A custom GPT is great for asking questions by hand. When you want the same analysis on every new export, without retyping prompts, move to Claude Code. It runs in your terminal and reads files on your machine, so you point it at the JSON the extension exported and describe the job in normal language.

You might say: "Read leads.json, flag every business with no website, score the rest for a local SEO pitch, and write the results to a scored.csv". Claude Code writes and runs the code, then hands you the file. Next week you drop in a fresh export and run the same instruction again. The details are in our guide on analyzing lead lists with Claude Code.

Whichever tool you pick, the raw material is identical: one clean export of real Google Maps businesses. That is what separates a useful lead analyst from a chatbot making things up, an idea we unpack further in feed your AI with business data.

When you want to act, not just analyze

Building a custom GPT is powerful for research and drafting. But a GPT does not remember which leads you called, does not plot them on a map, and does not send anything. When analysis needs to become a workflow, the leads flow into the Vonsel dashboard.

There the same businesses live inside a mapped CRM. Reviews Intelligence reads the review text and surfaces pain points. Email Intelligence writes an AI email tailored to each business. Instead of a flat CSV, you get a value added database where every lead carries its own context, ready to work. The data you fed your GPT and the data you act on stay in one place.

The businesses themselves come from public Google Maps listings, the same source the Google Maps Platform Places API exposes to developers, which is why the fields map so cleanly into any AI tool. If you want the legal background before you scale, read is it legal to scrape Google Maps.