free-lead-scraping

Find leads for free using Google search operators (dorks) and the Instant Data Scraper browser extension, then extract emails and phones from the export. Use when the user has no budget, no API keys,

SKILL.md

GitHub

Free Lead Scraping (System 3)

Goal

Get real contact details without spending anything. No API keys, no account, no credits.

When to use

  • The user has no budget or hasn't signed up for anything yet
  • They want to try the repo before paying for it
  • Anything they call System 3, dorks, or Instant Data Scraper
  • Suggest this first to anyone new — it's the only system that costs nothing, so it's the safest place to prove the pipeline works.

How it works

People put their email in their Instagram bio. Google indexed it. So rather than paying a database for contact details, ask Google for pages that already show them:

site:instagram.com "barber" "gmail.com" OR "yahoo.com" OR "outlook.com"

Process

Step 1 — build the queries

python -X utf8 scripts/03_free_scraping/build_dorks.py \
    --niche "barber" --location "Tuscaloosa AL" --output output/dorks.md

Writes a markdown file of ~12 query variants, each with a clickable Google link (pre-set to 100 results a page).

Step 2 — scrape the results page

Needs the Instant Data Scraper extension (free, Chrome or Brave):

  1. Run one of the queries.
  2. Click the extension. If the table preview looks wrong, click "Try another table".
  3. Click Locate "Next" button, then click Google's Next link.
  4. Click Start crawling, let it page through, then CSV.

Step 3 — pull the contacts out

python -X utf8 scripts/03_free_scraping/parse_scraper_export.py \
    --input ~/Downloads/instant-data-scraper.csv \
    --out-csv output/free_leads.csv

Regex only, so still free. It handles the usual obfuscation (name (at) domain (dot) com) and strips platform noise (@instagram.com, @wix.com, image URLs).

Add --llm to run Claude over rows the regex missed — costs about a tenth of a cent per row and needs ANTHROPIC_API_KEY. Worth it for free-text bios, not for structured pages.

Writing better queries

  • One city at a time. "barber Alabama" is noise; "barber Tuscaloosa" is businesses.
  • Use the word they use. Plumbers say "plumbing", lawyers say "attorney" or "law firm".
  • -site:yelp.com -site:yellowpages.com strips aggregators so you get the businesses.
  • Nothing returned? Drop the location first, then a provider. Too many quoted terms means no page contains all of them.
  • filetype:pdf is underrated — association member lists and sponsor sheets leak contacts constantly.

Edge cases

  • Google shows a CAPTCHA — you're paging too fast. Wait, or use a smaller &num=.
  • Everything got dropped — the parser discards rows with no email and no phone. Pass --keep-empty to see them.
  • Business names look wrong — the parser guesses from the longest short text cell. On messy pages it guesses badly; the emails are still right.

Cost

Zero.