How to Write a Website Chatbot Script That Qualifies Leads

Plenty of websites have a chat bubble in the corner. Far fewer have one that actually moves visitors toward a sale. The difference is rarely the technology. It is the script: what the bot asks, in what order, and what it does with the answers.

A good lead-qualifying chatbot does three jobs. It answers the easy questions instantly, it figures out whether this visitor is a fit, and it gets fit visitors to a next step, such as a booked call, while your team is busy or asleep.

This guide focuses on writing that script. We will use an invented business, Brightline Web Studio, a small agency that builds websites for local service companies, as the running example.

Start with the questions visitors already ask

Before you write a single greeting, collect real questions. Look at:

• Contact form submissions from the last few months
• Emails that start with “quick question”
• Notes from discovery calls
• Search terms from your site search, if you have one

Group them. For Brightline, the questions fell into five buckets: price, timeline, what is included, whether they work with a specific industry, and how to get started. Those buckets become the topics the bot must handle confidently. Anything outside them is a handoff.

If you use an AI chat agent rather than a rigid decision tree, feed it the same material as a knowledge base: service descriptions, typical price ranges you are comfortable sharing, service area, process and policies. The AI can then phrase answers naturally, but it still needs your facts.

Decide what “qualified” means before the bot does

A chatbot cannot qualify leads if you have not defined a qualified lead. Write down three to five criteria. For Brightline:

1. They run a service business (their niche).
2. They need a new site or a rebuild, not small edits.
3. Their budget falls within the studio’s range.
4. They want to start within the next three months.

Each criterion becomes one question in the script. Keep it to the minimum. Every extra question costs you visitors who drop off before sharing an email address.

The script, section by section

1. Opening line

Skip “How can I help you?” It invites a blank stare. Offer choices instead:

Hi, I’m the Brightline assistant. Are you looking for a new website, help with an existing one, or just checking prices?

Buttons for each option make the first reply effortless.

2. Answer first, then ask

If the visitor asks about price, answer before you qualify. Nothing kills trust faster than a bot that dodges a direct question.

Most of our service-business sites land between $4,000 and $9,000, depending on pages and features. What kind of business is the site for?

That second sentence starts qualification without feeling like a form.

3. Qualifying questions

Ask one question per message, in plain language:

• “Is this a brand-new site, or a rebuild of one you already have?”
• “Roughly when would you like it live?”
• “Do you have a budget range in mind? A ballpark is fine.”

4. Contact capture

Ask for contact details only after the visitor has received some value:

Thanks, that helps. What’s the best email or mobile number to send a short proposal outline to?

If you plan to text them, say so, and make sure your consent language and messaging registration are in order. Check current carrier and TCPA rules before sending automated texts.

5. Next step

For fit visitors, offer a booked call right inside the chat:

Looks like a great fit. Want to grab a 20-minute call with our project lead? Here are the next open times.

For visitors who are not a fit, stay helpful: point them to a resource or a partner, and still save their details if they agree.

6. Fallback

Write a fallback that admits the limit and offers a human:

I’m not sure about that one. I can have someone from the team reply by email within a business day, or you can call us now at (555) 010-0199.

Handoff rules: when the bot should step aside

Decide in advance when a person takes over. Common triggers:

• The visitor asks for a human.
• The same question fails twice.
• The lead scores as highly qualified and a team member is online.
• The topic is sensitive: a complaint, a billing dispute, a legal question.

With a live chat widget that supports AI handoff, the team member should see the full transcript and answers so far, so the visitor never repeats themselves. SaaSVisionary’s support desk chat widget, for example, lets AI cover routine questions and a person take over with the visitor’s context.

Route every conversation into the CRM

A chat that ends in a transcript nobody reads is wasted. Make sure each conversation:

1. Creates or updates a contact with the details captured.
2. Tags the contact with the qualifying answers (for example “rebuild”, “budget-mid”, “start-q1”).
3. Notifies the right person, by email, mobile app or team channel.
4. Starts a follow-up sequence if the visitor did not book.

When the chat lands in the same shared inbox as your texts and emails, whoever follows up sees the whole history in one thread.

Measure the script, then rewrite it

Track a handful of numbers monthly:

Metric	What it tells you
Chats started per 100 visitors	Whether the opening line earns a click
Share of chats reaching contact capture	Whether the questions are too many or too pushy
Qualified leads per week	Whether the bot finds real buyers
Calls booked from chat	Whether the next step is compelling
Top unanswered questions	Where the knowledge base has gaps

Read a sample of transcripts every week. They are the fastest way to spot a confusing question or a missing answer.

Frequently asked questions

How many questions should a lead qualification chatbot ask?

Usually three to five qualifying questions, plus contact details. Each question should map directly to a criterion you use to decide whether a lead is a fit, such as project type, budget or timeline. Beyond five, many visitors lose interest before sharing contact information. Ask one question per message and answer the visitor’s own questions first.

Should my chatbot share prices?

If you can give a realistic range, yes. Visitors who ask about price and get dodged often leave. A range such as “most projects fall between X and Y” answers the question honestly, sets expectations early and helps filter out leads whose budget is far from yours, which saves your sales team time.

What is the difference between a rule-based chatbot and an AI chatbot for lead capture?

A rule-based bot follows a fixed tree of buttons and scripted replies, which is predictable but rigid. An AI chatbot understands free-text questions and phrases answers naturally from your knowledge base. Many businesses combine both: AI for open questions, with structured steps for qualification and contact capture so the data you need is always collected.

When should a chatbot hand a conversation to a human?

Hand off when the visitor asks for a person, when the bot fails to answer the same question twice, when a highly qualified lead is ready to talk, or when the topic involves a complaint, a billing issue or anything sensitive. The team member should receive the full transcript so the visitor does not have to repeat anything.

Want to put a qualifying chat on your site? Start a free 14-day trial.