Score leads with ChatGPTFeed your AI the data, let it rank the list
Export your Google Maps leads as CSV, Excel, or JSON with the free Vonsel extension, then let ChatGPT and Claude Code score and prioritize them for you.
AI Workflows··7 min read
Key takeaways
The free Vonsel Chrome extension captures Google Maps businesses and exports them as CSV, Excel, or JSON. That file is the fuel for your AI.
ChatGPT scores a list in seconds when you upload a file and hand it a clear rubric. Claude Code turns that same rubric into a repeatable script for thousands of rows.
Scoring is only step one. The Vonsel dashboard takes ranked leads into a mapped CRM with review and email intelligence so the best rows become real conversations.
The idea
Feed your AI with real business data
Lead scoring used to mean building a spreadsheet formula or paying for an enrichment platform. Now you can hand a raw list to a model and ask it to rank the rows. The only thing you need first is clean, structured data, and that is exactly what a browser extension gives you.
The Vonsel extension runs on Google Maps, captures the businesses in a search (name, category, rating, review count, phone, website, address), and lets you export the whole set as CSV, Excel, or JSON. The scraping is the means. The end is a scored, prioritized list you can actually work.
Below is the practical flow: export the file, open ChatGPT or Claude Code, give it a rubric, and read back a ranked table. Most of this article stays on those two tools because they cover the vast majority of real use, with a short note on Codex and local models at the end.
Get the data first
Capture any Google Maps search and export it in one click. Then feed the file to your AI of choice.
Run a search on Google Maps ("dentists in Austin", "gyms in Berlin"), let the Vonsel pop-up capture the results, and export. Which format you pick depends on where the file is going next.
Format
Best for
Why
CSV
ChatGPT uploads
Universal. Every model and tool reads it without setup.
Excel
Manual review first
Easy to scan and clean rows by hand before scoring.
JSON
Claude Code and Codex
Structured fields are ideal when AI writes a script against them.
If you want the full reasoning on picking a format, our guide on JSON vs CSV for AI analysis breaks it down. For a step-by-step on getting the file out, see export leads as CSV, Excel, or JSON for AI. CSV itself follows a simple, well-defined shape described in RFC 4180, which is why models parse it so reliably.
Step 2
Score the list in ChatGPT
Open ChatGPT, attach your exported CSV or Excel file, and give it a rubric. The model's data analysis feature reads uploaded files directly, as described in the OpenAI file uploads FAQ. A prompt that works well looks like this:
"Here is a CSV of local businesses with columns for name, category, rating, review_count, has_website, and phone. Score each row from 0 to 100 as a sales lead for a web design agency. Reward businesses with no website, a decent rating, and at least 10 reviews. Penalize rows with no phone. Return a table sorted highest to lowest, with a one-line reason per business."
The key is giving the model your criteria, not just the file. Tell it what a good lead looks like for your offer and it will apply that rubric consistently across every row. You can iterate ("weight review count higher", "drop anything under 3.5 stars") and it re-scores instantly.
export formats from the extension: CSV, Excel, JSON
0
cost to download the extension: no trial, no card
1
rubric that scores your whole list at once
Step 3
Scale it up with Claude Code
ChatGPT is perfect for a one-off list. When your exports grow to hundreds or thousands of rows, or when you want the same scoring logic every single week, move to Claude Code. It runs in your terminal, reads your exported file, and writes a scoring script you can reuse.
A first prompt might be: "Read leads.json. Write a Python script that scores each business as a web design lead: +40 if has_website is false, +1 per review up to 30, +20 if rating is 4.0 or higher, minus 15 if phone is empty. Output a sorted CSV called scored_leads.csv with a score column and a reason column."
Claude Code writes the script, runs it, and shows you the ranked output. The difference from ChatGPT is durability: the logic is now saved as a file. Next week you export a fresh list from the extension, drop it in the same folder, and re-run. Your scoring becomes a repeatable pipeline instead of a chat you have to redo.
To go deeper on this, our walkthrough on analyzing leads with Claude Code covers cleaning, deduping, and enrichment in the same session.
Ready to fuel your AI?
Install the extension, capture a search, and export a file your model can score in seconds.
You have one list, want a fast answer, and are happy pasting or uploading the file. Great for testing a rubric before you commit to it.
Use Claude Code when
You score lists every week, handle thousands of rows, or want the logic saved as a script so every future export runs identically.
Codex also works
If you already live in Codex or a coding assistant, the same JSON export and the same rubric produce the same scored file. The data is portable.
Local LLMs for privacy
Running a local model keeps the list on your machine. Scoring is lighter than generation, so even a mid-size local model handles a rubric well.
A scored list is not the finish line. Ranking tells you who to contact first, but the value shows up when the top rows turn into personalized outreach and tracked follow-ups. That is the job of the platform behind the extension, not the spreadsheet.
From score to sale
Where scoring meets the Vonsel dashboard
Scoring a CSV in ChatGPT is a great start, but it leaves you with a flat file and a to-do list. The Vonsel dashboard is where a ranked list becomes real pipeline. Your captured businesses land in a mapped CRM, so you see the highest-scoring leads on a territory map instead of a spreadsheet row.
From there the platform layers on the parts a raw scrape cannot give you: Reviews Intelligence reads what customers actually complain about, Email Intelligence drafts a tailored message per business, and Sales Intelligence surfaces the buying signals worth acting on. The value-added database is contextualized with each lead's real pain points, not just a rating and a phone number.
You can also reopen a previous capture, add more businesses, and pull fresh reviews into leads you already saved, so your scored list keeps improving over time. For the bigger picture on turning raw exports into contextualized data, read feed your AI with business data and, once you are ready to write outreach, write cold emails with ChatGPT from scraped data.
The scrape gives you rows. Your AI ranks them, the dashboard closes them
Turn a search into a scored pipeline
Capture Google Maps, export CSV, Excel, or JSON, and score it with your AI. Then bring the winners into the mapped CRM. Explore features or see plans.
Yes. Upload a CSV or Excel file of leads exported from the free Vonsel extension, then give ChatGPT a clear rubric (rating, review count, category, whether a website exists). It returns a scored, ranked table you can sort and work from. For hundreds of rows, Claude Code runs the same scoring as a repeatable script.
What file format is best for scoring leads with AI?
CSV is the safest default because every model and tool reads it. Excel works well for people who want to review the file by hand first. JSON is the best fit for Claude Code and Codex when you want AI to write scripts against structured fields. The Vonsel extension exports all three for free.
Is it better to score leads in ChatGPT or Claude Code?
ChatGPT is faster for a one-off list you paste or upload. Claude Code is better when you want a reusable scoring script, want to process thousands of rows, or want the logic saved so every future export runs the same way. Most people start in ChatGPT and graduate to Claude Code as their lists grow.
Add Chrome Extension