Website Contact Finder - Email, Phone, Contact Form

Find the email address, phone number or contact form for any business website. Pay only when a contact is found.

Give it business websites from anywhere: a Google Maps scrape, a LinkedIn company list, a CSV, or another Actor's output. For each business it returns the best available way to get in touch.

Route	What you get
email	A real address published by the business, with a confidence level
form	A link to their contact form, when they publish no address
phone	A number from a tel: link
none	No route, plus the reason why

When there is no route you get a reason, and the five reasons are deliberately distinct because they call for different decisions:

Reason	What it means
no-contact-published	Site works, publishes no way to reach them
blocked-by-bot-protection	Served us a challenge page. Details may well exist
site-unreachable	Domain resolves but would not serve us. Often a block, not an outage
domain-does-not-resolve	No address record. The website is genuinely gone
no-mail-server	Domain cannot receive email at all

The middle three would be a single blank cell in any other tool. Keeping them apart matters: a live business that refuses datacenter traffic is still a good lead, and reporting it as a dead website would make you discard it.

This Actor detects bot-protection challenge pages and reports them. It does not try to defeat them.

Why the "no route" answer matters

Every other contact scraper returns an empty cell when it finds no email, and an empty cell tells you nothing. Roughly a quarter of businesses with no published email address still have a working contact form. They are reachable. The empty cell was hiding it.

And when there is genuinely no way in, the reason tells you what to do next rather than leaving you to guess. See the table above.

Coverage, measured and published

Nobody else in this category publishes a hit rate. Here is ours, with the sample described so you can judge whether it resembles your list.

Sample: 500 real businesses with websites, taken from OpenStreetMap across Manchester, Toronto, Dublin, Hamburg and Sydney. Offices, clinics and trades. Measured 2026-08-07. These 500 were held back and never used while building the Actor, so this is a fair test rather than a rehearsed one.

Route	Share
Email address	51.2%
Contact form	12.8%
Phone only	11.6%
No route found	24.4%
Reachable one way or another	75.6%

Coverage depends heavily on the market, so here is the split rather than one flattering average. A separate 300-business sample from New York, London, Melbourne and Berlin resolved 62%, against 75% here. Judge by the row that looks most like your list.

	businesses	email	form	phone	no route
.de	130	60.8%	12.3%	8.5%	18.5%
.ie	87	54.0%	9.2%	12.6%	24.1%
.com	145	50.3%	11.7%	13.8%	24.1%
.co.uk	12	50.0%	8.3%	0%	41.7%
.com.au	33	48.5%	18.2%	9.1%	24.2%
.ca	54	44.4%	20.4%	11.1%	24.1%

Quality of the email half, audited by hand:

• 99.6% of returned addresses were genuinely that business's (243 of 244, audited by hand on a separate 300-business sample)
• 99.1% reproduced exactly when the same businesses were scanned again
• 66% of the addresses in the 500-business run above were published as clickable mailto: links, which is the high-confidence tier. Set requireHighConfidence to take only those.

Contact form detection was audited separately: 95.9% of detected forms were genuine contact forms rather than newsletter signups, search boxes, blog comment boxes or login forms, all of which are explicitly rejected.

These numbers come from Western English and German speaking cities. Coverage in other markets is not yet measured, and we would rather say so than quote you a global figure we have not earned.

Pricing

You pay for outcomes, not for effort.

Event	Price
Run starts	$0.005
Email address found	$0.015
Contact form or phone found	$0.005
No route found	free

A list of 1,000 businesses costs about $7, and you can work that out before you press start. Tools that charge per page crawled cannot tell you what a run will cost until it has finished.

Records with no route are still returned, at no charge. You paid to find out.

Input

Three ways in. Use whichever fits.

From another Actor. Set datasetId to the upstream run's dataset. In an Actor-to-Actor integration that is {{resource.defaultDatasetId}}. The website column is detected automatically; override it with domainField if needed.

From a list. Put URLs in startUrls or bare domains in domains.

{
  "domains": ["outlandish.com", "grundmann-dentistry.de"],
  "maxPagesPerSite": 6,
  "requireHighConfidence": false
}

Output

{
  "domain": "grundmann-dentistry.de",
  "route": "email",
  "email": "info@grundmann-dentistry.de",
  "email_type": "general",
  "confidence": "high",
  "evidence": "mailto",
  "source_url": "https://grundmann-dentistry.de/kontakt",
  "form_url": null,
  "phones": ["+493012345678"],
  "all_emails": ["info@grundmann-dentistry.de"],
  "reason": null,
  "pages_crawled": 4
}

confidence is high when the business published the address as a mailto: link and medium when it was found in page text. Set requireHighConfidence to accept only the former, which is worth doing for cold outreach where bounce rate matters.

email_type tells you which door you are knocking on, and the Actor picks in this order:

Value	Meaning
personal	A named individual, such as j.smith@. Best for a personalised approach
general	The main inbox: info@, hello@, contact@
other	A word specific to that business, such as bookshop@ or skin@. Usually fine
department	A specific team: press@, careers@, advertising@. A real address, but often the wrong one to pitch

Addresses that can never receive a real message, such as noreply@ and postmaster@, are never returned as the contact. Filter on email_type if you only want general inboxes and named people.

What this Actor does not claim

It does not verify that mail will actually be delivered. It tells you the business published the address, which the 99.6% precision figure supports. Deliverability is a different claim requiring SMTP verification, and we do not make it.

Phone numbers come from tel: links only. Numbers written as plain text are not guessed at, because guessing them produces false positives.

Phone precision is lower than email precision, and here is exactly how. Audited across 462 numbers from 500 businesses:

• 75.3% come back as proper E.164 with a country code, for example +4940364520
• 23.8% are returned in national format without a country code. These come from .com and .org sites, where the domain gives no hint about the country. We return the digits as published rather than prefixing a + and inventing a country: a Toronto number written as +4169240785 reads as Swiss, and a confidently wrong number is worse than an obviously local one.
• 0.9% are discarded as unusable, mostly shortcodes and public service lines. 116117 was being returned for two separate German practices; it is the national out-of-hours medical number, not either practice's line.

One known gap: a number published on several unrelated businesses, usually from a shared template or agency, is not currently detected. It appeared on 3 of 500 businesses in testing.