Google Maps Scraper — Leads, Emails, Phone Numbers & Contacts

A Google Maps scraper built for lead lists. Give it search terms and locations and it returns one flat row per business — name, category, full address in parts, coordinates, phone, website, rating, review count, reviews per star, opening hours, claimed and closed status — and, if you ask, the emails and social links found on each business's own website. Use it as a Google Maps extractor for plain Google Maps data, as a source of Google Maps leads and local business leads, or as a Google Maps email extractor when you need Google Maps emails for outreach. It is priced the way reviewers of other actors keep asking for:

• $0.003 per place, details included. No add-on for details, no charge per filter, no start fee.
• $0.002 per contact — only when one is found. A place with no website, an unreachable website or a website with nothing on it costs $0 extra.
• Every filter is free and runs before billing. A place your filters reject is never charged. The status row tells you how many each filter removed.
• Charged equals delivered. Rows are charged first and only the charged rows are written, so an aborted run or a spending limit never leaves you paying for something you did not get.

Last verified working: 2026-09-21.

Google Maps scraper for AI agents: which input answers which request?

Built to be called by an AI agent (Claude, ChatGPT, or any MCP client through the Apify MCP server) as well as by people. One call, one flat table back, no API key for Google, no login, and the cost is known before the run: maxPlacesPerSearch × $0.003, plus $0.002 for each place where a contact is actually found.

Use this actor when the task is to find local businesses (any category, any country) and get their phone numbers, websites, addresses, ratings, opening hours, emails or social media contacts. Do not use it for review text, photos or popular times — it does not return them.

The user asks for	Input
"dentists in Austin with phone numbers"	{"searches": ["dentist"], "locations": ["Austin, TX"], "maxPlacesPerSearch": 50}
"…with emails" / "contact details" / "social media"	add "enrichContacts": true
"only businesses that have an email"	add "onlyWithEmail": true
"businesses without a website"	add "hasWebsite": "no", "hasPhone": "yes"
"unclaimed" Google Business Profiles	add "isClaimed": "no"
"low rated" / "bad reviews" businesses	add "maxRating": 3.9, "minReviews": 20
"top rated" / "established" businesses	add "minRating": 4.5, "minReviews": 100
"new businesses since last week" (scheduled)	add "onlyNew": true, "monitorKey": "my-list"
"every restaurant in the city", more than ~200 places	add "fullCoverage": true and raise maxPlacesPerSearch
"details / phone / email for this place" (URL, place ID or CID known)	{"startUrls": ["ChIJ…"], "enrichContacts": true}
several categories × several cities	"searches": ["plumber", "electrician"], "locations": ["Tulsa, OK", "Norman, OK"]

What comes back: rows with type: "place" (the data — name, category, phone, phone_international, website, address, city, postal_code, country_code, lat, lng, rating, review_count, opening_hours, place_id, google_maps_url, lead_flags, and with enrichment email_primary, emails, facebook, instagram, linkedin, whatsapp) and one free type: "status" row per search that says how many places were found, filtered, delivered and charged — read it to explain a short or empty result to the user instead of guessing. An unknown location or an empty search is a successful run that costs $0.

Google Maps scraper output: what is in a row?

One row per place, always the same columns in the same order, so CSV and Excel exports are never ragged. Fill rates were measured on 870 places from 12 cities in 6 countries on 2026-09-21.

Field	What it is	Filled
place_id, cid, feature_id, google_maps_url	The stable place id (ChIJ…), the CID from ?cid= links, the feature id, and a link that opens the place	100%
name, category, categories	Business name, primary category and every category, in the language you ask for	100%
search_term, search_location, rank	Which search returned the place and where it stood in the list	100%
address, street, city, state, postal_code, country_code, neighborhood	The one-line address and its parts. state and neighborhood exist only where the country uses them	100 / 99.8 / 91 / 66 / 91 / 100 / 76%
lat, lng, plus_code, distance_km_from_center, timezone	Coordinates, the global plus code, kilometres from the centre of your location or area, IANA timezone	100%
phone, phone_international	Local format and +1 512-… format	95%
website, domain	The website on the listing and its host	83%
rating, review_count, reviews_per_star	Stars, number of reviews, and the 1★–5★ histogram	100%
is_closed_permanently, is_closed_temporarily, is_claimed	Closed flags, and whether the listing has an owner (false = "Claim this business")	100%
opening_hours, open_now	{"monday": "8 AM–5 PM", …} and whether it was open when fetched	99%
description, attributes	The short description and the attributes listed as present (accessibility, amenities, service options)	31% / 86%
booking_url, menu_url, image_url, photo_count, price_level	Reservation link, menu link, cover photo, photo count. price_level and menu_url are in the schema but the search answer did not carry them in any of the 870 places — expect null	24 / 0 / 99.8 / 99 / 0%
has_website, has_email, lead_flags	Free lead helpers: lead_flags is any of no_website, low_rating (< 4.0), few_reviews (< 10), unclaimed	100%
fetched_at	UTC timestamp	100%

With enrichContacts on, every place row also carries: emails (each with type — named or role such as info@ — and the source_url it was read from), email_primary, phones_from_site, contact_page_url, facebook, instagram, linkedin, x, youtube, tiktok, whatsapp and enrichment_status.

A second row type, status, is written once per search and is always free: search_term, search_location, target, status (ok / no_results / unknown_location / error), requested, found, filtered, filtered_by (how many each filter removed), duplicates, already_seen, delivered, charged, contacts_found, pages, grid_cells, coverage_note, note and error. delivered always equals charged.

Google Maps leads by city: how do I get them?

Cross searches with locations. Each term is searched in every location; maxPlacesPerSearch caps what is delivered (and charged) per pair and maxPlacesTotal caps the whole run.

{
  "searches": ["dentist", "orthodontist"],
  "locations": ["Austin, TX", "Round Rock, TX"],
  "maxPlacesPerSearch": 100,
  "enrichContacts": true
}

A place that turns up under two searches is delivered and billed once — de-duplication is by place id across every search and grid cell in the run. A full phrase that already says where, such as "pubs near Prague 2", works as a search term on its own; leave locations empty.

language sets the language of categories, hours and descriptions (en, de, fr, pt-BR, ja …) and country the country to assume for ambiguous names (us, de, jp …). The release test covers the United States, Germany, France, Brazil, India, Japan and the UAE.

Google Maps emails: how does contact enrichment work, and what does it find?

Turn on enrichContacts. For each place with a website the actor reads the home page and then the most promising contact, about, team and imprint pages — up to maxPagesPerSite (default 5), because the address is usually not on the home page. It collects emails (including Cloudflare-protected and name [at] domain ones), tel: numbers, Facebook, Instagram, LinkedIn, X, YouTube, TikTok and WhatsApp links.

Measured on 200 places sampled across the 12 test cities on 2026-09-21:

Outcome	Places	Charged
Contact found (email or social link)	102 (51%)	$0.002 each
… of which with at least one email	67 (34% of all places, 41% of places that have a website)
Website read, nothing on it	21	$0
Website unreachable or refusing crawlers	25	$0
The "website" is itself a social profile	16	$0
No website on the listing	36	$0

So expect an email for roughly one place in three, not for every place — and pay for the ones you get. An address on somebody else's domain that is not even a mailto: link (the web agency's, a booking widget's, the template's) is left out rather than passed off as the business's email. Emails are not verified over SMTP.

The crawler identifies itself, honours robots.txt, reads at most one page per second per site, gives each page a 10-second timeout and reads HTML only.

onlyWithEmail delivers — and charges — only the places where an email was found. The run reads further down the result list to fill your limit.

Scrape Google Maps for businesses without a website

The prospecting list web and marketing agencies ask for: places with a phone but no website.

{
  "searches": ["plumber", "electrician", "roofer"],
  "locations": ["Tulsa, OK"],
  "hasWebsite": "no",
  "hasPhone": "yes",
  "skipClosed": true,
  "maxPlacesPerSearch": 50
}

hasWebsite, hasPhone and isClaimed take any, yes or no ("isClaimed": "no" keeps only listings Google still offers to "Claim this business"). The places that do have a website are counted under filtered_by.has_website in the status row and cost nothing.

Google Maps data for reputation work: how do I find low rated businesses?

{
  "searches": ["restaurant"],
  "locations": ["Lyon"],
  "language": "fr",
  "country": "fr",
  "maxRating": 3.9,
  "minReviews": 20,
  "maxPlacesPerSearch": 50
}

minRating / maxRating filter on stars and minReviews / maxReviews on the review count (0 = no limit). A place whose rating or count is unknown never passes a filter that needs it: nobody should pay for a row that cannot be shown to match. The Google Maps review count is on every row, reviews_per_star gives the 1★–5★ split, and lead_flags marks low_rating, few_reviews and unclaimed on every row for free.

More filters, all free: categoriesInclude / categoriesExclude match words against the place's own categories (the fix for "two thirds of my leads were off-industry and I was still charged"); nameContains / nameExcludes match the name; excludePlaceIds skips ids you already have and excludeDatasetId skips every place_id in an earlier dataset of this actor.

How do I monitor for new businesses on Google Maps?

{
  "searches": ["bakery"],
  "locations": ["Boulder, CO"],
  "maxPlacesPerSearch": 200,
  "onlyNew": true,
  "monitorKey": "boulder-bakeries"
}

With onlyNew, a run delivers and charges only places that earlier runs with the same monitorKey have not delivered; everything else is counted under already_seen. Schedule it weekly and each run is the delta. Only delivered places are remembered, so a place a spending limit cut off comes back next time. Use one monitorKey per list.

One honest caveat. Maps re-ranks and rotates the tail of every result list between requests: asked six times in a row, "bakery Boulder, CO" returned 111–129 places each time, 138 distinct, 90 in all six. The actor asks an exhausted list twice and merges the answers, which covers 93–99% of the union, but a repeat run can still surface a few places the previous run was never shown. They are new to you and are delivered as new; nothing is ever delivered twice.

How do I use my own Google Maps URLs, place ids or CIDs?

Paste them into startUrls: search URLs (google.com/maps/search/…, read from the map position in the URL), place URLs (google.com/maps/place/…), short links (maps.app.goo.gl/…), a Google Maps place id (ChIJ…), a feature id or a CID.

{
  "startUrls": [
    "ChIJb5gBC-u0RIYRHkntaUcdJGs",
    "https://maps.google.com/?cid=7720327853776914718",
    "https://www.google.com/maps/search/ramen/@35.658,139.7016,15z"
  ],
  "maxPlacesPerSearch": 40
}

A place gives one full row. The first two lines above are the same business pasted two ways — it is delivered once. An id Maps does not know is reported in a free status row.

If you are moving from another Google Maps actor, a saved input pastes unchanged: searchStringsArray, locationQuery, maxCrawledPlacesPerSearch and countryCode are accepted as aliases of searches, locations, maxPlacesPerSearch and country.

Searching an area: circle, bounding box or polygon

Instead of locations, give latitude and longitude with radiusKm, or a boundingBox, or a polygon:

{
  "searches": ["zahnarzt"],
  "latitude": "52.52",
  "longitude": "13.405",
  "radiusKm": 2,
  "withinRadiusOnly": true,
  "language": "de",
  "country": "de"
}

Maps ranks by prominence as well as distance, so a well-known place slightly outside your circle can appear. withinRadiusOnly drops anything outside the circle, box or polygon before billing — the answer to "results far outside my radius". distance_km_from_center is on every row either way. A boundingBox is {"south": …, "west": …, "north": …, "east": …}; a polygon is a list of [latitude, longitude] points.

FAQ: how many places can one search return, and what is full coverage?

A single Maps query lists a limited number of places and then stops — in our measurements between about 120 and 220, depending on the query. When that happens and you asked for more, the status row carries a coverage_note saying how many were listed.

fullCoverage goes past it: the area is split into a grid, every cell is searched separately from the centre outwards, and the results are merged. It works with locations (the area is looked up) or with a circle, box or polygon. gridCellKm sets the cell size (0 = automatic) and maxGridCells bounds the number of cells (default 64). In the release test a 3 km circle in central Austin returned 260 places for "cafe" through 9 cells, where single queries in the same city stopped between 203 and 222. Every place is still delivered and billed once, however many cells it appears in.

What happens with a small town, or a location that does not exist?

Every location is looked up first. If Maps does not know it, that search ends with status: unknown_location, a note, and no charge — and if it was the only search, the run still ends as a success with the explanation on the run card. This matters more than it sounds: asked for "plumber Zzqqxwv Nowhereville", Maps answers with twenty real plumbers near whatever IP asked, as if nothing were wrong. The actor never passes those off as your town.

A real small town never comes back as a silent zero either. If the typed query returns nothing, or returns places centred far from the looked-up town, the actor searches again around the town's centre and says so in the note.

What if Google throttles or blocks a request?

A consent page, a captcha or an empty shell is never read as "no more places". The request is retried on a fresh proxy session; if the search still fails it becomes an error row, not an empty success. There is one more thing the actor handles quietly: about half of all answers arrive with every review count stripped. Those are asked again, so review_count is filled on 100% of rows in the release test rather than on half of them.

Does it work outside the United States?

Yes, by default. The release gate runs 12 cities in the US, Germany, France, Brazil, India, Japan and the UAE and requires at least 60 places from 4 pages in each; on 2026-09-21 every city returned between 67 and 77.

Is the same search the same twice?

Mostly, and the README will not pretend otherwise: Maps re-ranks per request, so two identical runs capped at 80 places shared 68–77 of them in our tests. Set the cap above what Maps lists, or use fullCoverage, when you need the complete set rather than the top of it.

Is it legal to scrape Google Maps?

The actor reads publicly visible business listings — the same information anyone sees on the map without logging in — and the public pages of business websites, with robots.txt honoured. Business contact details are generally not personal data, but some listings belong to sole traders, and emails can be personal data under the GDPR and similar laws. You are responsible for having a lawful basis for how you use what you collect, for honouring opt-outs, and for following the anti-spam rules where your recipients live. If in doubt, ask a lawyer.

Pricing: what does a run cost?

Event	Price	When
place	$0.003	One place row delivered, all details included
contact-found	$0.002	Only when enrichment is on and an email or social link was read off that place's own website
Filters, status rows, duplicates, already-seen places, failed or empty searches, unknown locations, start	$0	Always

Worked example: 1,000 places with enrichment on, 380 contacts found → 1,000 × $0.003 + 380 × $0.002 = $3.76. Without enrichment the same list is $3.00.

Your spending limit is honoured to the row: when it is reached the run stops charging and stops delivering at the same moment, and a contact the limit refused is blanked rather than handed over unpaid. Aborting a run stops billing immediately, because nothing is charged until the moment it is written. There are no hidden free-tier caps.

Speed and proxy settings

Apify datacenter proxies work for Maps and are the default in proxyConfiguration; residential is not needed. sessions is the number of proxy sessions used in parallel (default 6) and perIp the requests per second per session (default 1). Measured on 2026-09-21: the 20-place example takes about 20 seconds, 260 places through a 9-cell grid about a minute, and 40 places with enrichment under two minutes — reading websites politely is the slow part. A filter that very few places pass (say, restaurants rated under 4.0) makes the run read the whole result list, which takes a couple of minutes per search.

Google Maps API alternative: run it from Python, JavaScript or curl

Python:

from apify_client import ApifyClient

client = ApifyClient("<APIFY_TOKEN>")
run = client.actor("kestrel/google-maps-scraper").call(run_input={
    "searches": ["coffee shop"], "locations": ["Austin, TX"], "maxPlacesPerSearch": 50, "enrichContacts": True})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
    if row["type"] == "place": print(row["name"], row["phone"], row["email_primary"])

JavaScript (Node):

import { ApifyClient } from 'apify-client';

const client = new ApifyClient({ token: '<APIFY_TOKEN>' });
const run = await client.actor('kestrel/google-maps-scraper').call({
  searches: ['coffee shop'], locations: ['Austin, TX'], maxPlacesPerSearch: 50, hasWebsite: 'no' });
const { items } = await client.dataset(run.defaultDatasetId).listItems();
console.log(items.filter((r) => r.type === 'place').map((r) => [r.name, r.phone]));

curl, synchronous, straight to CSV:

curl -X POST "https://api.apify.com/v2/acts/kestrel~google-maps-scraper/run-sync-get-dataset-items?token=<APIFY_TOKEN>&format=csv" \
  -H "Content-Type: application/json" \
  -d '{"searches":["coffee shop"],"locations":["Austin, TX"],"maxPlacesPerSearch":20}'

That last call is also the shortest route from Google Maps to CSV. Compared with the official Places API there is no per-field billing, no 60-result ceiling per query, and the emails and social links are not something that API offers at all — which is why people look for a Google Places API alternative for lead generation in the first place.

Google Maps lead generation in n8n, Make and AI agents

• n8n — use the Apify node (or an HTTP Request node with the curl call above): Run actor and get dataset → filter type = place → your CRM, Google Sheets or an email tool. With onlyNew and a schedule trigger, each execution passes on only the new leads. People searching for a Google Maps scraper n8n workflow usually want exactly this loop.
• Make — the Apify app's Run an Actor module followed by Get Dataset Items; map name, phone_international, email_primary and lead_flags into the next module.
• AI agents (MCP) — through the Apify MCP server an agent can call this actor as a tool: "find 30 dentists in Austin without a website" becomes one run with hasWebsite: "no". The flat row and the lead_flags column are there so a model does not have to interpret nested data.
• Zapier, Google Sheets, webhooks — any Apify integration works; the dataset exports as JSON, CSV, Excel or an RSS feed.

What this Google places scraper does not do

Said plainly, so nobody finds out from a bill:

• Review text. This actor returns the rating, the review count and the per-star histogram, not the reviews themselves.
• Photos and popular times. One cover image URL and the photo count only.
• Email verification. Addresses are read off websites; they are not checked over SMTP.
• Person-level leads. No names and job titles from bought databases — that data is the source of the worst reviews in this category. The Google Maps business data here is what the business itself publishes.
• Google Maps phone numbers for places that list none. About 5% of places have no phone on their listing; phones_from_site sometimes fills the gap when enrichment is on.

Support

Open an issue on the actor's Issues tab with the run id and what you expected; issues are answered personally and quickly, and a field you need is usually a small change. If a run fails because of the source rather than your input, it fails loudly on purpose — a daily canary runs this actor and checks both the places and the enrichment.