Analyze lead lists with ChatGPT Export your data, then feed your AI

Your AI is only as good as the data you feed it

Large language models are excellent at reasoning over a list, but they do not have your leads. Ask ChatGPT to "find dentists in Chicago with weak websites" and it will guess, hallucinate names, or refuse. It has no live view of Google Maps. What it does have is a very sharp ability to analyze a structured file you hand it.

That is the whole workflow: capture real business data first, then feed your AI. The Vonsel extension scrapes the businesses straight from a Google Maps search, and exports a clean file with names, categories, addresses, phone numbers, websites, ratings, and review counts. The extension is the means. Your AI does the thinking.

This post walks the practical flow end to end, mostly through ChatGPT and Claude Code, with a short note on Codex and local models at the end. No coding required for the ChatGPT path.

Capture and export the list

Open Google Maps, run a search like "hair salons in Austin", and let the Vonsel pop-up capture the results. When the list looks right, export it. The extension gives you three formats, and the one you pick depends on which AI you plan to use.

Every row is real data pulled from the map, so your AI is reasoning over facts instead of inventing them. For a deeper look at the export options, see exporting leads as CSV, Excel, or JSON for AI and the step-by-step guide to export Google Maps data to CSV. The CSV format itself follows a simple standard documented in RFC 4180, which is why every AI tool reads it the same way.

Upload the file to ChatGPT and prompt it

This is the fastest path and the one most people start with. In a ChatGPT conversation, click the attach button, add your CSV or Excel file, and describe what you want. ChatGPT reads the columns, holds them in context, and answers in plain language or a fresh downloadable table.

The trick is to prompt against the exact columns Vonsel exports. Here are prompts that work well on a Google Maps lead file. See our fuller list of ChatGPT prompts for lead lists for more.

Because ChatGPT runs the numbers on the actual file, you can trust the counts in a way you never could with a made-up answer. For scoring specifically, our guide on scoring leads with ChatGPT goes deeper, and uploading a lead CSV to ChatGPT covers the mechanics. If you want to understand the philosophy behind all of this, read feed your AI with business data.

Claude Code for big lists and repeatable pipelines

ChatGPT in the browser is perfect for a few hundred rows. Once you are working with thousands of businesses, or you want to run the same analysis every week, Claude Code is the better tool. It reads the exported file directly from your folder, writes a scoring script, runs it, and hands you a clean result. Nothing to paste.

Export JSON from the extension so the fields stay structured, drop it in a folder, and open Claude Code there. A prompt like the ones below does the heavy lifting.

The advantage is repeatability. Claude Code keeps the script, so next week you export a fresh capture, run the same command, and get an updated analysis in seconds. For a focused walkthrough, see analyzing leads with Claude Code.

Codex and local LLMs

If you live in a terminal, OpenAI Codex works much like Claude Code: point it at the JSON file and describe the transformation you want. It is a natural fit when your lead analysis is one step in a larger script that already runs in your stack.

Privacy-sensitive lists are a good case for a local model. Running a model on your own machine (through a runner like Ollama or LM Studio) means the lead file never leaves your laptop. Local models are slower and less capable than the frontier options, but for straightforward tasks like tagging categories or flagging missing fields they hold up well. Export JSON, feed it in, and keep everything on-device.

Whatever you reach for, the input is the same clean file the extension produced. That is the point of feed your AI with data: the capture step is shared, and you swap the model to fit the job.

Where the leads actually live

Analysis is the middle of the story, not the end. Once ChatGPT has scored your list and Claude Code has cleaned it, you still need somewhere to work those leads: track status, contact them, and follow up. That is where the exported CSV starts to feel thin.

The Vonsel dashboard is built for exactly that. The same businesses you captured land in a mapped CRM, plotted on GPS so you can see your territory. Its Reviews Intelligence reads each business's reviews for pain points, its Email Intelligence drafts a tailored email per business, and its Sales Intelligence surfaces the leads worth calling first. You get value-added, contextualized data instead of a flat spreadsheet.

So the AI workflow and the dashboard are not rivals. Feed ChatGPT the file when you want quick analysis, and let the dashboard hold the relationship over time. If you want the wider context on using this kind of data responsibly, our note on whether it is legal to scrape Google Maps is worth a read, and you can browse more ideas across the AI Workflows blog.