Scrape any business in Spain then feed the data to your AI

The map is the source, your AI is the engine

Spain has a dense, well-mapped local economy. Restaurants in Barcelona, dentists in Madrid, workshops in Valencia, agencies in Bilbao: nearly all of them live on Google Maps with a name, an address, a phone, a website, and a pile of reviews. That public surface is a goldmine if you can pull it into a file and hand it to a model.

The workflow is simple. First you capture the businesses with a browser extension. Then you export them. Then you feed your AI. The extension is just the means to an end: getting structured Spanish business data out of the map and into a place where a language model can reason over it.

This guide walks the full loop. Most of it is about the AI half, because that is where the leverage is. Anyone can download a CSV. The advantage comes from what you do with it next.

Capture Spanish businesses from Google Maps

Open Google Maps and run a query the way you always do: a category and a place. Try business leads across Spain, or narrow it to something like restaurantes en Barcelona or clinicas dentales en Madrid. The extension pop-up appears on the map and starts capturing what Google shows.

For each listing you get the fields that matter for outreach: name, full address, phone, website, category, rating, and review count. The extension also enriches websites in the background to pull emails, phones, and social profiles when they are published. Everything you need to feed your AI with business data is captured in one pass.

Spain runs on the same public map data that Google Maps Platform exposes, so the fields you capture are consistent whether you are scraping a village in Galicia or the center of Sevilla.

Export the file that feeds your AI

Once the list is captured, export it. The format you pick depends on the tool you plan to hand it to, and each one has a natural home.

If you want the deeper reasoning behind picking one over another, our guide on exporting leads as CSV, Excel, or JSON for AI breaks it down. As a rule of thumb: CSV for chat, JSON for code.

Work the list in ChatGPT

ChatGPT is the fastest way to get value from a fresh capture, because you talk to it in plain language. Upload the CSV, then prompt. There is no setup, no code, and no import wizard. Our walkthrough on uploading a lead CSV to ChatGPT covers the mechanics, but the prompts below are where it earns its keep.

Segment. "Here is a CSV of dentists in Madrid. Group them into three tiers by rating and review count, and tell me which tier is worth calling first." In seconds you have a prioritized list instead of a flat spreadsheet.

Score. "Add a column that scores each business from 1 to 10 as a web design prospect. Weight low ratings, few reviews, and a missing website higher." Now the file tells you where the easy conversations are.

Draft. "For the top 10 rows, write a short cold email in Spanish that references one specific detail from their profile." That is outreach copy grounded in real data, not a generic template. If cold email is your goal, the companion piece on writing cold emails with ChatGPT from scraped data goes further.

ChatGPT shines for interactive, one-off analysis: you see a result, you ask a follow up, you refine. It is a conversation with your Spanish lead list.

Scale it with Claude Code

When one CSV becomes ten, or when you want the same analysis every week, a chat window stops scaling. This is where Claude Code takes over. Export the capture as JSON, drop it in a folder, and point Claude Code at it in your terminal.

Because it writes and runs code, Claude Code turns your Spanish business data into a small, repeatable pipeline. A few things it does well:

The mental model: ChatGPT is the analyst you chat with, Claude Code is the engineer that builds you a tool. For a deeper dive, see analyzing leads with Claude Code.

Codex and local LLMs

ChatGPT and Claude Code cover most cases, but the same JSON export feeds other engines too. Coding assistants like Codex are handy when you want a quick throwaway script to reshape a file, join two captures, or convert JSON to a format your CRM expects.

If the data is sensitive or you simply prefer to keep it on your machine, a local LLM running through a tool like Ollama or LM Studio can read the same export offline. You lose some raw capability, you gain full privacy. Either way the pattern is identical: the extension produces the file, the model does the thinking.

Skip the plumbing with the Vonsel dashboard

The raw-file-plus-AI loop is powerful, but it is still plumbing you maintain. If you would rather have the intelligence built in, that is exactly what the Vonsel dashboard is for. It is the end that the extension is the means to.

Inside the dashboard your captured Spanish businesses land in a mapped CRM. Reviews Intelligence reads the review text and surfaces pain points per business. Email Intelligence drafts a tailored AI email for each one. Sales Intelligence scores and prioritizes. You can reopen a previous scrape to add more businesses, attach reviews to captures you already made, and keep a history of everything you pulled.

In other words, the value-added, contextualized database you would otherwise build by hand with ChatGPT and Claude Code is already assembled for you, grounded in the real pains hidden in each business profile.