Website Email & Contact Extractor + AI (Lead Finder)

Turn a list of company websites into a clean contact list. This actor crawls each site's homepage plus its most contact‑relevant pages (contact / about / team / imprint) over plain HTTP — no API key and no browser — and pulls out:

• Emails (from mailto: links and visible text, with asset/placeholder junk filtered out)
• Phone numbers (only from real tel: links, so you never get garbage from page markup)
• Social profiles — LinkedIn, X/Twitter, Facebook, Instagram, YouTube
• A split of role/generic vs personal emails, on‑domain addresses first

Then the optional built‑in AI reads what was found and, for each site, picks the single best outreach contact, guesses the likely departments, writes a one‑line company summary, and drafts a personalized opening line tailored to what you sell.

Input

Field	Type	Description
websites	array (required)	Company websites/domains, one per line (acme.com or https://acme.com).
maxPagesPerSite	integer	Pages to crawl per site (homepage + most contact‑relevant pages). Default 4, max 10.
aiEnrichment	boolean	Pick the best contact, departments, summary and opener with AI. Default true. Turn off for raw extraction only (no AI charges).
yourOffer	string	What you sell, so the AI tailors the opener (e.g. "recruiting software for tech teams"). Optional.
aiLanguage	string	Language for the AI‑written fields. Default English.
proxyConfiguration	object	Apify Proxy is recommended (rotating IPs).

Example input

{
  "websites": ["cleveroad.com", "thoughtbot.com"],
  "maxPagesPerSite": 4,
  "aiEnrichment": true,
  "yourOffer": "recruiting software for tech teams",
  "aiLanguage": "English"
}

Output

One dataset item per website, for example:

{
  "website": "cleveroad.com",
  "reachable": true,
  "contactsFound": true,
  "pagesScanned": 4,
  "emailCount": 5,
  "emails": ["sales@cleveroad.com", "..."],
  "roleEmails": ["sales@cleveroad.com"],
  "personalEmails": [],
  "phones": ["+1 218 4131"],
  "linkedin": "https://www.linkedin.com/company/cleveroad",
  "twitter": "https://twitter.com/CleveroadInc",
  "bestContactEmail": "sales@cleveroad.com",
  "aiQualification": {
    "bestContactEmail": "sales@cleveroad.com",
    "bestContactReason": "Primary sales inbox for new business enquiries.",
    "likelyDepartments": ["Sales", "Business Development"],
    "companyOneLiner": "Cleveroad is a software development company building web and mobile products.",
    "personalizedOpener": "Hi Cleveroad team — loved how you frame your delivery process...",
    "outreachAngle": "Faster technical hiring to staff your project pipeline."
  }
}

If a site can't be reached you still get a row with reachable: false and the reason.

How it works

1. Fetches the homepage with rotating proxies and realistic headers.
2. Strips <script>, <style>, <svg>, <noscript> and <template> before reading text — so SVG path data and inline scripts can never be mis‑read as phone numbers or emails.
3. Extracts emails (mailto: + cleaned text), phones (tel: only) and social links.
4. Follows up to a handful of same‑site contact/about/team pages for more.
5. (Optional) Calls the built‑in AI once per site that has contacts, to qualify the lead.

Notes & limits

• Sites that hide all contact details behind a JavaScript form or a captcha may return no emails — that's expected for a browserless, respectful crawler.
• The actor only visits pages on the same domain you provide.
• Please use the output in line with applicable anti‑spam and data‑protection laws (e.g. GDPR/CAN‑SPAM). You are responsible for how you contact the people you find.