Official Website Finder: Bulk Company Name to Domain Lookup

Official website finder for lists of company names. Paste names, get back each company's official domain and website, a confidence label, and a short note when the match is weak. It is a company name to domain and company name to website lookup built on public sources only: Wikidata (open data, CC0) plus a DNS and homepage check. No search engine is scraped.

You are charged only for rows labelled high or medium. Names that are not found, or only found with low confidence, are free.

What does the bulk domain finder return?

For every name you enter you get one row, in the same order:

Field	Meaning
inputName	The name as you typed it
matchedName	The Wikidata label or the site name the match is based on
domain	Registered domain of the official site (no www, no subdomain)
website	Homepage URL after redirects
confidence, confidenceScore	high, medium, low or none, and the 0 to 1 score behind it
method	wikidata, dns-homepage or none
verified	The homepage was read and the company name appears in its title, site name or headings
title	Homepage title
wikidataId	Wikidata Q-number when the method is wikidata
mxFound	The domain has MX (mail) records: true, false, or null (not checked or DNS failed)
contactPageUrl	Contact page linked from the homepage (optional)
country	Country from Wikidata, or your country hint
note, error	Why a row is weak, ambiguous, blocked or unusable
charged, checkedAt	Whether the row was billed, and when it was checked

How to find a company website from a name

1. Open the Actor and enter one company name per line (legal suffixes such as Inc, LLC or GmbH are fine).
2. Leave the defaults on, or switch on Find the contact page if you also want contactPageUrl.
3. Run it. Export the dataset as JSON, CSV or Excel, or call it from your own code.

Input example

{
  "companies": ["Stripe", "Notion Labs", "Basecamp", "Linear", "Patagonia", "Hims & Hers"],
  "country": "",
  "verify": true,
  "checkMail": true,
  "includeContactPage": false,
  "maxCompanies": 500,
  "concurrency": 6
}

Output example

{
  "inputIndex": 0,
  "inputName": "Stripe",
  "matchedName": "Stripe",
  "domain": "stripe.com",
  "website": "https://stripe.com/",
  "confidence": "high",
  "confidenceScore": 1.0,
  "method": "wikidata",
  "verified": true,
  "title": "Stripe | Financial Infrastructure to Grow Your Revenue",
  "wikidataId": "Q7624104",
  "mxFound": true,
  "contactPageUrl": null,
  "country": "United States",
  "note": null,
  "error": null,
  "charged": true,
  "checkedAt": "2026-10-25T12:00:00Z"
}

How it works

1. Wikidata. The name is searched in Wikidata. Candidates must be organisations (company, business, brand, enterprise and similar classes, or a company-like description), must have an official website, and their label or alias must match your name. Family names, places, songs and software "by" someone are rejected. The official website is the best "official website" (P856) statement, not ended and not deprecated.
2. DNS and homepage fallback. When Wikidata gives nothing confident, the Actor guesses domains from the name (for example hims-and-hers, himshers on .com, .io, .co, .ai, .app, .org, .net, plus your country's ending), keeps the ones that resolve in DNS, opens their homepage and scores how well the page matches the name. Parked or for-sale pages, and social or directory sites, are rejected.
3. Verification. The homepage of the chosen domain is read (robots.txt respected, one page per site) and the name is looked for in the site name, page title, headings and text.

How accurate is it?

Honest numbers from a local test on 64 company names (famous brands, small businesses, non-US companies and deliberately ambiguous names). The thresholds were tuned on the first 32; the other 32 were written down before the first run and then used for three small rule changes, so treat the second set as a mostly clean check, not a guarantee.

Confidence	First set (32, tuned)	Second set (32, mostly held out)
high	24 of 24 correct	19 of 21 correct
medium	6 of 6 correct	3 of 4 correct
low	not billed	not billed

Across both sets, 52 of 55 high and medium rows were right (94 percent), and 52 of the 60 names that have a known answer came back as high or medium (87 percent). Your own list will differ, especially for small local businesses. What the numbers do not tell you:

• Wikidata covers companies that have a Wikipedia-level footprint: large companies, well-known brands, funded start-ups and many notable small brands. A neighbourhood plumber or a bakery is usually not in it.
• Small businesses depend on the DNS and homepage fallback, which only finds a company whose domain is close to its name and whose homepage shows the name in the title or headings. Expect many local businesses to come back none or low. That is intentional: a wrong domain is worse than no domain.
• Ambiguous names return low confidence or the wrong company. "Linear" (the product-management app) was found, but a name like "Mercury", "Apple Records" or "Intercom" can point to a different organisation with the same name. When two organisations tie, the best one is marked down and the note says so. Add a legal suffix, a city or a country hint if you can.
• Some big sites block automated requests (HTTP 403 or 429). Their Wikidata match is kept, verified stays false and the note says why. The Actor does not try to get around bot protection.
• Wikidata can be out of date or have a different domain from the one you expect (a brand site instead of the corporate site, or the other way round).

How confidence is computed

• Wikidata match: 0.45 for the name match (identical, or the same apart from words such as "Labs", "Group", "Inc"), plus 0.20 when Wikidata's instance-of class says organisation (0.10 when only the description does), plus up to 0.15 when the domain name equals the brand. A verified homepage adds 0.20 (0.15 for a weaker match in headings or the first word of the name). A homepage that does not mention the name takes 0.10 away. A name that only matches the start of the Wikidata label, or a tie between organisations with different domains, is capped below medium.
• DNS and homepage match: up to 0.40 when the site name or a title segment equals the company name (less for a partial match), plus 0.30 when the domain equals the name, plus small bonuses for the heading, meta tags and .com or your country ending. Parked pages score zero. If two guessed domains both match, the result is capped at low.
• Labels: high is 0.80 or more, medium 0.60 or more, low 0.35 or more, otherwise none. If the homepage cannot be read, a guessed domain is never medium or high.

How much does it cost?

You pay $0.70 per 1,000 domains found (the domain-found event, only for rows with confidence high or medium). Rows with confidence low or none, rows with an input error, and duplicate names in the same list are free. Platform usage is small: a typical row needs about 2 Wikidata requests and 3 to 4 page requests (robots.txt plus the homepage). Set Max companies to cap spending.

Use cases

• Add a website column to a CRM export or lead list that only has company names.
• Turn a list of names into domains before running an email or tech-stack lookup.
• Check that supplier or investor lists point to the real company sites, with an MX flag to see which domains receive email.

FAQ

Does it use Google, Bing or another search engine? No. It uses the Wikidata API and the company's own homepage only.

Is the data legal to use? Wikidata is released under CC0. Homepages are read with a plain request, one page per site, with robots.txt honoured. No login or bot protection is bypassed.

Why do I get none for a company I know exists? It is probably not in Wikidata and its domain does not look like its name, or its homepage blocks automated requests. Those rows are free.

What does verified: false mean on a high row? Wikidata matched strongly but the homepage could not be read (blocked, robots.txt, timeout) or does not show the name. The note field gives the reason.

Can I send a domain instead of a name? The Actor is for names. If a name looks like a domain (stripe.com), only the brand part is used.

Does the contact page option crawl the site? No. It reads the homepage you already checked and returns the contact link it finds there; it is null when the homepage has none.

Use with AI agents / MCP

Call the Actor from an AI agent through the Apify MCP server or the Apify API: pass companies as a list of names and read the dataset rows. An agent can trust high rows, ask a human to confirm medium rows, and treat low or none as "not found". The confidence, method, verified and note fields explain every decision, so an agent does not need to guess.

Related tools

• Contact Page Finder: finds the contact page and form of a website in depth.
• Work Email Finder: Find Email by Name & Domain (No SMTP): next step once you have the domain.
• Company Enrichment Tool & API: a fuller profile of a company from its domain.