# How to use your website to train AI lead scoring
> Use your website URL to help AI define your target customer and score LinkedIn prospects based on how well they actually fit your business.
- HTML: https://getkinlead.com/ai-lead-scoring-from-website
- Markdown: https://getkinlead.com/ai-lead-scoring-from-website.md

**Short answer:** You provide a website URL to an AI agent that analyzes your landing page and case studies to build a target persona. The AI then scores LinkedIn prospects from 0 to 100 based on how closely their title and industry match the customer described on your site.

## Identify buying signals on your website

AI doesn't just read words, it looks for patterns of who actually pays you. If your site says you help companies grow, that's too vague for a machine. If it says you help 50 person agencies scale outbound, the AI has a specific set of attributes to find on LinkedIn.

The best signals are usually in the sections where you prove your value. Testimonials and case studies work better than the homepage because they name the roles and company types that got results. If every case study features a VP of Marketing at a fintech company, that's your target.

Look for specific nouns and numbers. Phrases like 'for Series A startups' or 'designed for Head of Operations' are high-signal. These tell the AI exactly which filters to apply to a LinkedIn profile to determine if the person is a match.

Avoid using generic marketing slogans as your primary source. Phrases like 'empowering businesses to reach their potential' provide no data for a scoring model. The AI needs to know the size of the business, the industry, and the specific job function of the person who signs the check.

| | Website Section | What the AI Extracts | LinkedIn Equivalent |
| --- | --- | --- | --- |
| Value Proposition | The specific problem you solve | Industry and job function |
| Customer Logos | Company size and sector | Company headcount and industry |
| Case Studies | Job titles of successful users | Current role and seniority |
| Pricing Page | Target budget and scale | Company revenue or employee count |

## Convert website content into a lead profile

Turning a URL into a qualification rule means moving from marketing language to strict constraints. You have to define who the customer is and who they are not. This stops the AI from scoring every person in a general industry as a high fit.

Extract the commonalities from your best customers. If your site mentions fast growing startups, define that in numbers. Decide if that means 11 to 50 employees or 51 to 200 so the AI doesn't have to guess.

A lead profile is a set of weights. For example, if your website emphasizes 'Enterprise Security', a profile with 'CISO' in the title should get a massive score boost, while a 'Junior Analyst' at the same company should get a lower score despite the company match.

Consider the geographic constraints mentioned on your site. If your 'Contact' page only lists a New York office and your case studies are all US-based, the AI should be told to penalize profiles from other regions to avoid wasting your outreach limits.

1. **Extract core traits.** Copy text from your About and Features pages into a document. Ask an AI to find the three most common traits of the customers mentioned.
2. **Define negative constraints.** List who your website does not serve. If you only work with US companies, the AI should score any non-US profile as zero.
3. **Build a mandatory checklist.** Create five non-negotiable criteria. For example, the lead must be a Head of Sales in the SaaS industry at a company with 50+ employees.
4. **Map keywords to roles.** List every job title that appears in your case studies. Use these as the primary anchors for the AI to identify high-fit leads.

## Set your scoring thresholds

A score from 0 to 100 only works if you know where to draw the line. You shouldn't treat a 60 and a 90 the same. High scores are for people who match your current offer perfectly. Medium scores are for people who might fit but need more discovery.

The goal is to automate the no so you only spend time on the yes. If a lead doesn't match your mandatory criteria, they should be low fit regardless of how impressive their title is.

Think of your thresholds as a filter for your energy. A 90+ score means you can use a highly personalized, direct approach. A 70 score might require a softer intro to see if they actually have the problem your website describes.

Avoid the temptation to lower your threshold just to get more leads. A high volume of low-score leads leads to lower reply rates and increases the risk of your account being flagged for irrelevant outreach.

**Vague scoring**

- 80-100: Looks like a good lead
- 50-79: Maybe a lead
- 0-49: Not a lead

**Attribute-Based scoring**

- 80-100: Matches job title, industry, and company size exactly as described on the site
- 50-79: Matches industry and seniority but has a slightly different job title
- 0-49: Works in the right industry but is too junior or at a company size the site doesn't target

## Test the profile against real LinkedIn data

Don't trust a scoring model until you test it against a known sample. You need to see if the AI's interpretation of your website matches your actual experience with customers.

Pick ten profiles. Five should be people you'd pay to get on a call with today, and five should be clear bad fits. If the AI scores a bad fit as an 80, your website content is too vague or your constraints aren't strict enough.

Pay close attention to the 'false positives'. These are the leads that look good on paper but are actually useless. If you see a pattern, like the AI scoring consultants as high fit when you only want founders, add 'Consultant' to your negative constraints.

Testing is an iterative process. You will likely need to adjust your website's wording or the AI's instructions three or four times before the scoring feels natural.

1. **Create a test set.** Gather five perfect customer profiles and five non-customer profiles from LinkedIn.
2. **Run the scoring logic.** Pass these profiles through your AI scoring rules to see the numbers.
3. **Adjust the criteria.** If perfect leads score low, add more keywords from your website to the positive list. If bad leads score high, add them to negative constraints.
4. **Verify alignment.** Repeat this until the scores match your manual judgment for all ten profiles.

## Refine your website for better AI training

If the AI is struggling to score leads accurately, the problem is often the source data. Your website is the training manual for the AI. If the manual is vague, the output will be vague.

Add a 'Who we help' section to your landing page. Be explicit. Instead of 'We help B2B companies', write 'We help B2B SaaS companies with 20 to 100 employees who are struggling with churn'. This gives the AI a clear rubric.

Use specific industry terminology. If you serve the 'Logistics' sector, use words like 'freight forwarding', 'last-mile delivery', and '3PL'. The AI uses these terms to find a deeper match than just the broad 'Transportation' category on LinkedIn.

Review your case studies for title consistency. If one case study mentions a 'Growth Lead' and another says 'VP of Marketing', the AI might see them as different personas. Standardize the language to help the AI group these roles together.

```
Who we help:
- Company size: 11-50 employees
- Industry: Fintech, Insurtech, Neobanks
- Roles: Head of Compliance, Chief Risk Officer, VP of Legal
- Geography: North America and UK
- Pain point: Manual KYC processes taking more than 48 hours
```

## Doing this with Kinlead

Kinlead automates this process. Instead of manually extracting traits and building checklists, you provide your website URL. Kinlead finds people on LinkedIn who match your site and scores each lead from 0 to 100 against that profile.

Leads below your fit line are not kept, which keeps your pipeline clean. You can organize these scored leads into [lead groups](/features/lead-groups-and-scoring) for campaigns or export them to CSV.

The system uses LinkedIn's standard search, so you don't need a Sales Navigator seat to start scoring. Once you have a group of high-score leads, you can run [AI outreach](/features/ai-linkedin-outreach) that uses your sales playbook to book meetings.

> Kinlead automates the scoring process based on your website's value proposition so you only target high-fit prospects.

## Related

- [AI lead scoring framework](https://getkinlead.com/ai-linkedin-lead-scoring-framework.md)
- [Find leads](https://getkinlead.com/tools/find-leads.md)
- [Lead groups and scoring](https://getkinlead.com/features/lead-groups-and-scoring.md)

## Common questions

**Can I use a blog post instead of a homepage for lead scoring?**

Yes, if the post describes a specific customer problem and solution more clearly than your homepage. The AI uses the personas mentioned in the post to score leads.

**Does the AI need a Sales Navigator account to score leads?**

No, lead search and scoring use standard LinkedIn search parameters.

**How often should I update the website URL used for training?**

Update it when you change your pricing, target industry, or core value proposition so the AI scores against your current business model.

**What happens if my website is too vague for the AI to find a pattern?**

The AI will likely give a wide range of medium scores. You should then manually add specific job titles and company sizes to the lead profile to tighten qualification.

[Create a Kinlead account](https://getkinlead.com/signup) or [book a call](https://getkinlead.com/demo) to talk through pricing and setup.
