How to automate LinkedIn lead sourcing

Fresh account = instant ban on automation. Warming is baseline.

• Week 1: profile setup (photo, headline, 3-5 jobs history), 5-10 connection requests/day to real people in industry, like 10-20 posts daily
• Week 2: 15-20 connection requests/day, comments on posts (2-3 daily), 1 own post
• Week 3: 20-30 connections, 5+ comments, own posts 2-3/week, DMs to new connections
• Week 4: stabilization at safe max — can start automation

Warming is MANUAL — there is no point automating warming, that defeats the purpose. 15 min/day for 28 days.

LinkedIn tracks browser fingerprint. Each account MUST have a stable, unique fingerprint throughout its lifetime. We use playwright-extra + custom fingerprint:

// fingerprint/store.ts
import { chromium } from 'playwright-extra';
import stealth from 'puppeteer-extra-plugin-stealth';
chromium.use(stealth());

const accountFingerprints = {
  'account-01': {
    userAgent: 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) ...',
    viewport: { width: 1440, height: 900 },
    locale: 'en-US',
    timezoneId: 'Europe/London',
    proxy: 'http://user:pass@proxy-uk-1.smartproxy.com:7000',
  },
  // ... per account
};

export async function launchForAccount(accountId) {
  const fp = accountFingerprints[accountId];
  const browser = await chromium.launch({
    headless: true,
    proxy: { server: fp.proxy },
  });
  const context = await browser.newContext({
    userAgent: fp.userAgent,
    viewport: fp.viewport,
    locale: fp.locale,
    timezoneId: fp.timezoneId,
    storageState: `./storage/${accountId}.json`, // persistent cookies
  });
  return { browser, context };
}

storageState holds session cookies — so you do not log in every time (logging in is a trigger).

Sales Navigator search → profile list. Each action has a rate limit. Safe thresholds (2026):

• Profile views: 80-100/day (300+ = ban warning)
• Connection requests: 80-100/week (changed since 2023, was 100/day)
• Search queries: 30-50/day
• Message sends: 30-50/day to connections

const RATE_LIMITS = {
  profileViews: { max: 80, window: '24h' },
  connectionRequests: { max: 80, window: '7d' },
  searchQueries: { max: 30, window: '24h' },
};

async function checkAndScrapeProfile(accountId, profileUrl) {
  if (!await canPerform(accountId, 'profileViews')) {
    logger.info({ accountId, action: 'skip', reason: 'rate_limit' });
    return null;
  }
  // ... actual scrape
  await recordAction(accountId, 'profileViews');
}

Distribute work across multiple accounts — 3 accounts × 80 profiles/day = 240/day without risk.

A bot doing page.goto → page.locator.click → next every 2 seconds = obvious bot. We simulate a human:

async function humanVisit(page, url) {
  await page.goto(url, { waitUntil: 'domcontentloaded' });
  await sleep(rand(1500, 3500)); // initial pause
  // Random scroll
  for (let i = 0; i < rand(2, 5); i++) {
    await page.evaluate(`window.scrollBy(0, ${rand(200, 600)})`);
    await sleep(rand(800, 2000));
  }
  // Random mouse movement
  await page.mouse.move(rand(100, 1300), rand(100, 700), { steps: rand(10, 25) });
  await sleep(rand(500, 1500));
  // Reading time on profile
  await sleep(rand(8000, 25000)); // 8-25s per profile
}

Per profile: 15-40 seconds (vs <2s for a naive bot). 80 profiles daily takes 30-60 min wall-clock. That is OK — automation does not have to be instant.

Raw profiles = noise. Filter through LLM to keep only ICP fit:

import Anthropic from '@anthropic-ai/sdk';
const client = new Anthropic();

async function scoreProfile(profile) {
  const response = await client.messages.create({
    model: 'claude-3-5-sonnet-latest',
    max_tokens: 256,
    messages: [{
      role: 'user',
      content: `ICP: B2B SaaS, 50-200 employees, founder/CRO/Head of Sales.
        Profile:
        ${JSON.stringify(profile)}
        
        Score 0-10 for ICP fit. Reply with JSON: {"score": N, "reason": "..."}`,
    }],
  });
  return JSON.parse(response.content[0].text);
}

Push to HubSpot only score >=7. HubSpot API:

async function pushToHubSpot(profile, score) {
  await fetch('https://api.hubapi.com/crm/v3/objects/contacts', {
    method: 'POST',
    headers: {
      'Authorization': `Bearer ${process.env.HUBSPOT_TOKEN}`,
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({
      properties: {
        firstname: profile.firstName,
        lastname: profile.lastName,
        company: profile.company,
        jobtitle: profile.title,
        linkedin_url: profile.url,
        ax_score: score.score,
        ax_score_reason: score.reason,
        lifecyclestage: 'lead',
      }
    }),
  });
}