Company Website Classifier — industry, B2B/B2C & ICP fit

Part of LeoWorks lead and hiring data tools: places → leads (Google Maps Lead Qualifier) · company websites → ICP fit (Company Website Classifier) · job postings → filters (Job Posting Classifier).

For sales teams, lead-gen agencies, B2B marketers and AI agents that already crawl company websites — and need to classify company websites — what each company does, who it sells to (B2B or B2C) and ICP fit scoring against their ideal customer profile, for $2 per 1,000 websites, with no prompt writing or LLM key.

• Industry — 16 groups: software, finance, healthcare, consumer goods & retail, food & hospitality, fitness, manufacturing, distribution & logistics, construction & real estate, professional services, marketing, media, education, energy, government & nonprofit, other
• Business type — software / online store / marketplace / in-person services / professional services / manufacturer / distributor / financial services / media / nonprofit & public
• Customers — businesses (B2B) / consumers (B2C) / both
• Reach — local / regional / national / international
• ICP fit score (0–100) — describe the companies you want in plain language (e.g. "Independent dental practices, not chains" or "B2B SaaS companies that sell to ecommerce brands")
• Your own labels — up to 10 yes/no criteria (e.g. "offers a free trial", "mentions HIPAA compliance")
• Tier and reason — A/B/C/D and a one-line reason per company, for CRM notes
• Contacts found on the site — email (the company's own domain first), phone, LinkedIn, Facebook, Instagram, X and YouTube, from the page text, links and schema.org data — no extra charge
• Pricing, careers and contact pages — links when the crawl found them
• No wasted spend — parked or for-sale domains, Facebook/Linktree-style pages and sites with too little text are skipped and not charged

Use it to: turn a crawl of hundreds of websites into a ranked lead list · drop chains, vendors and wrong-fit companies before outreach · segment a market by industry and business model · check which prospects publish pricing · feed labelled companies to a CRM, a dashboard or an AI agent.

How this differs from our other lead tools: this one reads the company's own website — what it says it does, sells and serves. Our Google Maps Lead Qualifier judges places from their Google Maps listing instead; use both to go from a Maps search to qualified websites.

How to use

1. Crawl the companies' websites with Website Content Crawler (or RAG Web Browser) — one start URL per company, Max crawling depth 1, about 5 pages per site.
2. Open this Actor, pick that crawl's dataset, and describe your ideal customer in ICP (optional).
3. Optional: up to 10 Custom labels, Only websites with an email, and the start URLs in Expected websites to see which sites the crawl missed.
4. Click Start. 6 websites (56 pages) take about 3 seconds.
5. Download COMPANIES.csv (best fit first, with tier, reason and contacts) or open the Companies and Contacts views.

From a Google Maps search to qualified websites

1. Score places with our Google Maps Lead Qualifier. Its QUALIFIED_WEBSITES record is a ready Website Content Crawler input: home pages only, no social pages, depth 1, about 6 pages per site.
2. Run Website Content Crawler with it.
3. Run this Actor on the crawl with the same ICP and the start URLs in Expected websites. Join the two results on domain.

Tested end to end on 2026-10-11: 60 Chicago places → 7 qualified websites → Website Content Crawler (56 pages, $0.008) → this Actor co17MPWY0GBRXshGL: 6 companies classified, 1 site the crawl could not read reported as missing.

Output sample

Real rows from run b8pJYBhGWLkwiS8GG (2026-10-10): 211 websites crawled with Website Content Crawler, ICP "Independent dental practices (not chains)", one custom label ("offers emergency or same-day appointments").

website	fit	industry	business type	customers	reach	custom: same-day
greentreedentalchicago.com	89	healthcare	in-person services	consumers	local	yes (0.94)
themeadowsdentalcare.com	88	healthcare	in-person services	consumers	local	yes (0.98)
kidstoothteam.com	72	healthcare	in-person services	consumers	regional	no
nexhealth.com (dental software)	36	software	software	businesses	national	yes (0.52)
heartland.com (dental support organization)	23	healthcare	professional services	businesses	national	no
dental360usa.com (25 clinics)	7	healthcare	in-person services	consumers	regional	yes
aspendental.com (national chain)	4	healthcare	in-person services	consumers	national	yes
gorgias.com	2	software	software	businesses	international	no

Each row also has the homepage url and title, pagesFound / pagesUsed, the page language, pricingPage / careersPage links, and the probability and confidence of every label.

Input example

The form default — two pasted homepages, no dataset needed (about $0.004, 2 seconds):

{
  "pages": [
    { "url": "https://acme-analytics.example.com/", "text": "Acme Analytics helps mid-size retailers forecast demand. … plans from $49 per month." },
    { "url": "https://brightsmile-dental.example.org/", "text": "Bright Smile Dental is a family-owned dental practice in South Austin. …" }
  ],
  "icp": "Independent dental practices (not chains)"
}

To classify a crawl, pass its dataset instead — pages are grouped by website automatically:

{
  "datasetId": "YOUR_WEBSITE_CONTENT_CRAWLER_DATASET_ID",
  "icp": "B2B software companies that sell to ecommerce brands",
  "customLabels": ["offers a free trial or free plan"]
}

Crawling tips: start Website Content Crawler with the homepages you want (one per company). Max crawling depth 1 is enough — the Actor uses the homepage plus up to 4 of the most telling pages (about, products or services, pricing, customers). Its default browser crawler reads sites built with JavaScript; the faster "Raw HTTP" crawler is cheaper but misses some of them (reported as skipped or missing, not charged). Website Content Crawler removes page footers by default, where social links usually are; links in schema.org data are still found.

Pricing

Pay only for classified websites — no subscription.

Event	Price	When
company-classified	$0.002	One website (company) classified from its crawled pages: industry, business type, customer type, reach, ICP fit and any custom labels. Websites with too little text are not charged.

That is $2 per 1,000 websites, whatever the number of pages crawled per site. First run with the form defaults: about $0.004 (2 websites, 2 seconds).

Cost examples

Websites	Cost
100	$0.20
1,000	$2.00
10,000	$20.00
50,000	$100.00

With the free $5 monthly Apify credit you can classify about 2,500 websites.

Websites with too little text (under 80 characters, usually sites that need a browser to render) or only bot-check pages are listed with a skipped reason and not charged. Websites that fail after retries are reported with an error field and not charged. If you set a maximum cost per run, the Actor stops cleanly when it is reached.

Works with

Source on Apify Store	Page URL	Text	Notes
Website Content Crawler (apify)	url	text / markdown	title, meta description, language, depth, Open Graph site name and schema.org data used — checked against real output (2026-10-10, 2026-10-11)
RAG Web Browser (apify)	metadata.url / searchResult.url	markdown	published schema
Contact Details Scraper (vdrmota)	url	—	its emails, phones, linkedIns, facebooks, instagrams are merged into the contacts; add its pages to a crawl dataset for the text
Other crawlers (current_url, start_url, pageTitle, top-level description, language, depth)	✓	text / markdown / content	published schemas

If none of the items has a page URL the Actor recognises, the run stops with a message instead of finishing with 0 websites. Also: the Apify API and JavaScript/Python clients · Apify Schedules · Claude, Cursor and Claude Code through the Apify MCP server (below).

Call it from code

curl -X POST "https://api.apify.com/v2/acts/leoworks~company-website-classifier/run-sync-get-dataset-items?token=YOUR_APIFY_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"datasetId": "YOUR_CRAWL_DATASET_ID", "icp": "B2B SaaS companies that sell to ecommerce brands"}'

JavaScript (npm install apify-client):

import { ApifyClient } from 'apify-client';
const client = new ApifyClient({ token: 'YOUR_APIFY_TOKEN' });
const run = await client.actor('leoworks/company-website-classifier').call({ datasetId: 'YOUR_CRAWL_DATASET_ID', icp: 'Independent dental practices (not chains)' });
const csv = await client.keyValueStore(run.defaultKeyValueStoreId).getRecord('COMPANIES.csv');

Python (pip install apify-client):

from apify_client import ApifyClient
client = ApifyClient("YOUR_APIFY_TOKEN")
run = client.actor("leoworks/company-website-classifier").call(run_input={"datasetId": "YOUR_CRAWL_DATASET_ID", "icp": "Independent dental practices (not chains)"})
fits = [c for c in client.dataset(run["defaultDatasetId"]).list_items().items if c.get("tier") in ("A", "B")]

Use with Claude, Cursor or Claude Code (MCP)

Ask your AI agent. Add the Apify MCP server with this Actor as a tool, then say, for example:

• "Does this company fit “B2B software companies selling to retailers”? https://acme-analytics.example.com/ — “Acme Analytics helps mid-size retailers forecast demand. Connect Shopify, NetSuite or your POS, and get store-level forecasts every morning. Trusted by 300 retail brands in the US, UK and Canada.”"
• "Classify the websites in my Website Content Crawler dataset YOUR_DATASET_ID and keep only the companies that fit “Independent dental practices (not chains)”, with the fit score and reason."
• "Tell me the industry, business type, B2B or B2C and reach of this company from its homepage text: “Bright Smile Dental is a family-owned dental practice in South Austin. Dr. Lee and our team offer cleanings, fillings, Invisalign and same-day crowns. New patients welcome.”"

The agent calls the tool leoworks--company-website-classifier and reads the labels with get-dataset-items. Replace YOUR_DATASET_ID with the ID of your own dataset.

Claude Desktop or Cursor (mcp.json):

{
  "mcpServers": {
    "apify": {
      "url": "https://mcp.apify.com?tools=leoworks/company-website-classifier",
      "headers": { "Authorization": "Bearer YOUR_APIFY_TOKEN" }
    }
  }
}

Claude Code: claude mcp add --transport http apify "https://mcp.apify.com?tools=leoworks/company-website-classifier" --header "Authorization: Bearer YOUR_APIFY_TOKEN". Leave out the header to sign in with OAuth in the browser instead. We verified this setup with the Apify MCP server (v0.17.4) on 2026-10-10: the agent classified a pasted homepage against an ICP in 9 seconds end to end (run x8Ykm1pLEJkM3f8a9).

Output

Where	What
Dataset — view Companies	One row per website: company name, fit score, tier, reason, industry, business type, customers, reach, email, phone, pricing and careers pages
Dataset — view Contacts	Email(s), phone, LinkedIn, Facebook, Instagram, X, YouTube, contact page
Dataset — view Labels with probabilities	Every label with probability and confidence
COMPANIES.csv	All classified companies, best ICP fit first, with rank, reason and contacts
REPORT (JSON)	Label shares across all companies, ICP match rate and the 25 best fits

One row

From run co17MPWY0GBRXshGL (2026-10-11, build 0.2.2; labels shortened; the company name and contacts are replaced by fictitious values):

{
  "domain": "brightfit-studio.example",
  "companyName": "BrightFit Studio",
  "url": "https://brightfit-studio.example/",
  "title": "BrightFit Studio – Personal Training and Functional Fitness",
  "email": "hello@brightfit-studio.example",
  "phone": "773-555-0142",
  "fitScore": 94,
  "tier": "A",
  "qualified": true,
  "industry": "fitness_wellness",
  "businessType": "in_person_services",
  "customerType": "consumers",
  "reach": "local",
  "reason": "Fits the ICP (94/100, tier A). Fitness & wellness, in-person services. Sells to consumers. Local reach. Contact found: email, phone.",
  "labels": { "icpMatch": { "probability": 0.94, "matched": true } }
}

fitScore is the ICP match probability × 100 and qualified is true from ICP match threshold (0.5). Without an ICP, rows have the four company labels only. A narrow, two-part ICP ("SaaS that sells to retailers") gives lower scores than a one-part one — sort by fitScore and lower the threshold if too few companies qualify.

Summary (REPORT)

Each run also saves a REPORT record (Output tab → Summary by label) at no extra charge: the count and share of every industry, business type, customer type and reach, the share of websites with a pricing page, and — with an ICP — how many matched plus the 25 best-fitting websites.

{
  "companies": 203,
  "icp": "Independent dental practices (not chains)",
  "icpMatched": 14,
  "topFits": [{ "domain": "smile-dental.example", "fitScore": 96, "industry": "healthcare", "url": "https://www.smile-dental.example/" }],
  "pricingPageShare": 0.118,
  "industry": { "healthcare": { "count": 33, "share": 0.163 }, "software_internet": { "count": 31, "share": 0.153 } }
}

Accuracy

Measured on 179 hand-labelled English company websites crawled with Website Content Crawler (2026-10-10): 62 employers from a sample of US job postings, 37 local businesses from Google Maps (dentists, gyms, restaurants) and 80 companies chosen to cover SaaS, online stores, agencies, manufacturers, distributors, finance, media and nonprofits. Each website was also paired with two of ten ICPs — its own kind and a near miss (e.g. a dental chain or dental software against "independent dental practices"). The questions were adjusted on one part, then checked once on a separate part that was not used for adjusting.

Set	Websites	Industry	Business type	Customers	Reach	All four right	ICP pairs	ICP fit right
Check set — not used for adjusting	109	99%	98%	98%	97%	93%	218	99%
Adjusting set	70	100%	100%	100%	96%	96%	140	99%

A label counts as right when it matches the hand label or an alternative noted for companies that fit two answers (e.g. Stripe is both software and financial services; a company in the US and Canada is national or international). ICP pairs that a careful reader could answer either way (e.g. an insurer against "banks, lenders and fintech") count as right either way — 24 of the 218 check pairs. Of the check-set websites that truly fit their ICP, 97% were found. Websites with only a few lines of text (often sites that need a browser crawler) give the model little to go on — check those rows. Automated labels can be wrong; check samples before making big decisions.

Limits

Item	Limit
Pages used per website	Homepage (first 3,000 characters) + up to 4 more pages (1,500 characters each)
ICP	Up to 1,000 characters
Custom labels	Up to 10, each up to 200 characters
Dataset size	Any number of pages; all pages are read first, then each website is classified once
Language	English (measured). Other languages are accepted and usually work, but accuracy has not been measured — treat them as beta
Speed	About 200 websites in 6 seconds
Contacts	Up to 5 of each kind per website, from crawled pages only — sites that hide them in images or scripts give none
Data	Only page URLs, titles, meta descriptions and page text are sent for classification. Contact details are read from your crawl by code and never sent to the AI; use them in line with privacy and anti-spam law

FAQ

Which AI makes the judgments? Jev, TypeSafe's decision model (version jev-1.13.0, pinned). Jev answers each label with a calibrated probability instead of generated text, so a fit score of 80 means the same thing across websites; the same site can move a few points between runs.

Does it crawl websites? No. It classifies pages you already crawled — run Website Content Crawler first (see Works with) or paste page text.

How is this different from tools that take a list of domains? Those crawl each domain themselves and charge for it. This one uses a crawl you already have (or a cheap one you run with depth 1), so you pay $2 per 1,000 companies for the judgment — and the crawl can be reused for other things. Add Website Content Crawler's cost to compare (about $0.008 for 7 sites with the Raw HTTP crawler in our test).

Which websites are skipped and not charged? Parked, for-sale or "coming soon" domains, pages on social or directory hosts (facebook.com/yourshop, linktr.ee), sites whose crawl has too little text, and — with Only websites with an email — sites without an email.

Why is a website counted once if I crawled 50 of its pages? You pay per company, not per page. The Actor picks the pages that say most about the company (homepage, about, products or services, pricing, customers).

Disclaimer

Independent tool — not affiliated with, endorsed by or sponsored by the companies shown or by the author of the crawler listed above. Company names are used only as examples of public websites.

Reviews and support

If this Actor saved you time, a short review on Apify Store helps others find it. Questions or a dataset whose fields are not detected? Open an issue in the Issues tab — we answer within a day.

Changelog

See the Changelog tab.