Google Maps Scraper (No proxies, no paid APIs)

A multi-worker Google Maps business scraper with a desktop GUI. No proxies, no paid APIs — drives real Chromium browser profiles with human-like behavior to collect business data at scale into CSV and SQLite.

Features

• Multi-worker scraping — run up to 20 parallel browser workers, each with its own persistent profile and isolated location queue
• No proxies required — relies entirely on stealth/anti-detection techniques rather than IP rotation
• Anti-detection layer
  • Removes navigator.webdriver and fakes plugins / Chrome runtime / WebGL fingerprints
  • Human-like mouse movement (bezier curves, easing, micro-jitter)
  • Human-like scrolling with overshoot and easing
  • Gaussian-distributed delays instead of uniform random waits
  • Random idle/"reading" pauses
  • CAPTCHA / block detection with automatic cooldown and retry
  • Rotating user-agent and viewport per worker (Windows + macOS Chrome/Edge)
• Resumable — already-scraped places (by URL) are skipped on re-runs, per keyword + location
• Live desktop GUI (CustomTkinter, dark theme)
  • Per-worker status panel (idle / starting / running / done) with progress
  • Live scrollable log with color-coded message types
  • Running totals: records collected, scrape rate, elapsed time, ETA
  • Start/stop controls that finish the in-flight record cleanly
• Built-in location database — a bundled offline dataset of 250 countries, 5,300+ states, 152,970 cities and 844,248 postcodes (from dr5hn/countries-states-cities-database). Click Browse locations to add a single city, an entire state, an entire country (all cities, or the top N by population), or postcodes/ZIP codes — no CSV needed.
• Bulk locations — import a CSV of City/State/Country (with column-mapping dialog), or add locations manually as chips
• Filters — skip businesses that don't match your criteria before they're written:
  • Min/max review count
  • Min/max rating
  • Must / must not have a website
  • Must / must not have a phone number
• Website classification — each result is tagged as Legit Website, Social Media Page, Yellow Pages Link, or No Website
• Optional extras — opening hours, weekly schedule, and recent reviews (configurable depth)
• Dual output
  • Append-only CSV per run (UTF-8 with BOM, opens cleanly in Excel)
  • SQLite database (WAL mode) per keyword, for dedup and querying
• Config persistence — last-used keyword, depth, worker count, and review settings are saved to scraper_config.json

Requirements

• Windows 10/11
• Python 3.10+
• Playwright (Chromium)
• CustomTkinter

Installation

git clone https://github.com/athm793/local-business-scraper.git
cd local-business-scraper

pip install -r requirements.txt
playwright install chromium

Optional: desktop shortcut

py create_shortcut.py

Creates a "Maps Scraper" shortcut on your Desktop that launches the GUI with pythonw (no console window).

Usage

GUI (recommended)

Double-click MapsScraper.pyw (or the desktop shortcut) to launch with no console window. From a terminal:

pythonw MapsScraper.pyw

py gui.py also works and additionally opens a console for debugging (set MAPS_SCRAPER_CONSOLE=1 to keep it; otherwise gui.py re-launches itself under pythonw so no terminal lingers).

Use the ⤓ Tray button (top toolbar) to minimize the app to the system tray; right-click the tray icon for Show / Quit.

1. Enter a search keyword (e.g. "dentist").
2. Set results per location and parallel workers.
3. Optionally configure filters (review count, rating, website, phone) and reviews/hours/schedule options.
4. Add locations — click Browse locations to pick from the bundled database (single city, whole state/country, or postcodes), type one manually, or upload a CSV (columns: City, State/Province, Country). On first use the picker unpacks the bundled database once (~250 MB decompressed, one-time, offline).
5. Click Start. Watch live per-worker status, totals, and the log panel.
6. Click Copy path or Open folder to grab the resulting CSV from outputs/.

CLI (single location)

py scraper.py --keyword "dentist" --location "Miami, FL" --depth 100

Output

CSV (outputs/<keyword>_<timestamp>.csv)

One row per business with columns:

keyword, location, city, state, country,
name, category, address, phone, website, website_type,
rating, review_count, hours, schedule,
latitude, longitude, place_url, scraped_at, reviews

SQLite (<keyword>.db)

A businesses table (WAL mode, place_url unique) used for dedup across runs — re-running the same keyword/location skips places already scraped.

Project structure

gui.py            # App entry point / window assembly
gui_widgets.py     # Theme constants + reusable widgets (cards, chips, dialogs)
gui_config.py      # Load/save scraper_config.json
gui_locations.py   # Location CSV import + manual entry
gui_runner.py       # Start/stop, queue polling, live stats
gui_picker.py       # Location picker dialog (browse the bundled world database)
geodb.py            # Read-only access to the bundled countries/states/cities/postcodes DB
pool.py            # Async worker pool — N browser contexts, shared queue
scraper.py         # Core Playwright scraping logic + CLI entry point
stealth.py         # Anti-detection: fingerprinting, mouse/scroll, block handling
db.py              # SQLite schema + upsert/dedup
csv_writer.py      # Thread-safe append-only CSV writer
tray.py             # System-tray (minimize to tray, Show/Quit)
MapsScraper.pyw     # No-console launcher (pythonw entry point)
create_shortcut.py # Windows desktop shortcut generator
data/              # Bundled world database (world.sqlite3.gz, decompressed on first run)

Location data

The Browse locations picker is powered by dr5hn/countries-states-cities-database (release v3.2-export.7), bundled as data/world.sqlite3.gz and decompressed to data/world.sqlite3 on first use. That database is licensed under the Open Database License (ODbL); it remains under ODbL here — see data/ATTRIBUTION.txt. The scraper's own code is MIT (below).

Notes

• Each worker launches a real, visible (non-headless) Chromium window with its own persistent profile under browser_profile_<n>/. Headless mode is intentionally disabled — it's far more detectable.
• Worker launches are staggered (8s apart) to avoid simultaneous requests from the same machine.
• This tool automates a public website. Use it responsibly and in accordance with Google's Terms of Service and applicable laws in your jurisdiction.

Contributing

Issues and pull requests are welcome. For significant changes, open an issue first to discuss what you'd like to change.

License

MIT