How to Train Your Chatbot

Your chatbot is only as good as the documents you give it. A well-trained chatbot answers prospect questions accurately, captures leads, and gives you pre-call intelligence. A poorly-trained one gives vague answers and drives visitors away.

This guide covers what to upload, how to format it, and how to keep your chatbot sharp over time. If you haven't set up your chatbot yet, start with the AI Chatbot Setup guide first.

Minimum Requirements

To get useful results from your chatbot, you need at minimum:

• 3-5 documents covering your core topics
• ~1,500 words of total content (enough for the chatbot to have substance to draw from)
• The basics covered: what you do, who you help, and how much it costs
• At least one FAQ document (Q&A format matches how visitors ask questions)

The chatbot will not perform well with a single short document. Think of it like briefing a new team member - the more context you give them, the better they perform in conversations with prospects.

What to Upload

Start with these high-impact documents first:

Add these over time as your chatbot matures:

• Technical documentation for evaluators who want to dig deeper
• Onboarding guides that show new customers what to expect
• Industry-specific use cases for different verticals
• Whitepapers or research that establish thought leadership

You don't need all of these on day one. Start with the top 5 and expand based on what your Knowledge Gap reports tell you prospects are asking about.

How to Format Documents

How you write your documents matters as much as what you include. These formatting guidelines help the AI find and serve the right answers.

Use Q&A format wherever possible

Write content as questions and answers. When a prospect types "How much does this cost?", a document with "Q: How much does your product cost?" is a near-perfect match.

Good:

Q: How much does your product cost?

Our pricing starts at $49/month for individuals and $29/seat/month for teams of 5+. All plans include...

Less effective:

A paragraph buried in a brochure that mentions pricing in passing.

If you already have marketing content that isn't in Q&A format, create a companion FAQ document that covers the same topics as questions and answers. You don't need to rewrite everything - just add the Q&A layer.

Use clear headings and sections

Break documents into sections with descriptive headings. Each section should cover one topic. Avoid generic headings like "More Info" or "Details" - use descriptive headings like "Pricing and Plans" or "How Our Onboarding Works."

Use concrete numbers and specifics

Specific details retrieve better than vague language:

Put the most important answer first

Lead with the direct answer, then add supporting details. The AI pulls the most relevant chunk of your document - make sure the answer is at the top of each section, not buried in the third paragraph.

Use your prospects' language

If your customers say "cost" and "budget", use those words - not just "pricing" and "investment." Think about how your prospects ask questions in real sales conversations and mirror that language. If you're not sure what terms they use, check your chatbot conversation logs - prospects will tell you in their own words.

Write lead-in sentences before lists

Instead of a bare bullet list, add a sentence explaining what the list covers. This gives the AI context for what follows and improves retrieval accuracy.

Describe visual content in text

The chatbot cannot read images in your PDFs. If a diagram, chart, or screenshot contains important information, describe it in the accompanying text. A table of pricing tiers in text format works - a screenshot of a pricing page does not.

Common Mistakes

One massive document instead of several focused ones

A 50-page PDF retrieves worse than 5 focused documents of 5-10 pages each. The AI chunks documents into sections - smaller, topic-focused documents mean cleaner chunks and better answers. If you have a large document, split it by topic: pricing in one file, product features in another, case studies in a third.

Uploading marketing brochures as-is

Brochures are designed for humans scanning visually. They have sparse text, lots of whitespace, and information spread across layouts the AI can't parse well. Convert the key content into text-based Q&A format instead.

Outdated or contradictory information

If your pricing changed last month but your uploaded docs still show old prices, the chatbot will give wrong answers. Keep documents current - outdated content actively hurts credibility with prospects.

Missing the obvious questions

Upload documents that answer what prospects actually ask - not what you wish they would ask. Check your chatbot's Knowledge Gap report to see what questions go unanswered, then create content to fill those gaps.

Duplicate content across documents

If three documents all explain your pricing differently, the chatbot may retrieve conflicting information. Keep each topic in one authoritative document. It's fine to reference pricing briefly in a product overview, but the detailed breakdown should live in one place.

Test Your Chatbot

After uploading documents, test with real prospect questions before sharing your profile widely:

1. Ask your top 10 prospect questions - The ones you hear on every sales call
2. Check for accuracy - Are the answers specific, with real numbers? Or vague and generic?
3. Try different wording - Ask the same question multiple ways. Does it still find the right answer?
4. Test the boundaries - Ask something NOT in your documents. The chatbot should say it doesn't have that information and suggest contacting you directly.
5. Verify pricing - Ask about pricing and check it returns exact figures
6. Try edge cases - Misspellings, very short questions ("pricing?"), questions that span multiple topics

If the chatbot gives vague or wrong answers, the fix is almost always better documents - not a different setting. Go back to the formatting guidelines above and check whether your content follows them.

Keep Your Chatbot Sharp

Check Knowledge Gaps

Your chatbot dashboard shows questions it couldn't answer well. These are direct signals from prospects telling you what content to create next. A few minutes each week reviewing gaps keeps your chatbot improving.

Go to Chatbot > Analytics to see your knowledge gap rate and unanswered topics.

Update documents when things change

New pricing? New feature? New case study? Upload the updated document. Remove or replace the outdated version so the chatbot doesn't serve stale information. A good habit is to review your uploaded documents whenever you update your website or marketing materials.

Review conversation topics monthly

Topic distribution in your dashboard shows what prospects care about most. If 40% of questions are about pricing, make sure your pricing document is comprehensive. If a topic is trending up, add more depth there.

Go to Chatbot > Conversations to see topic breakdowns and conversation trends.

Add seasonal or event-specific content

Attending a conference? Upload a document covering your booth offerings, event-specific promotions, or relevant use cases for that audience. Remove it after the event if it's no longer relevant. Launching a new product? Add documentation before the launch so your chatbot is ready when traffic arrives.

Quick Reference Checklist

Use this checklist when setting up or auditing your chatbot's knowledge base:

• At least 3-5 documents uploaded
• Minimum ~1,500 words of total content
• At least one FAQ-style document included
• Pricing with concrete numbers included
• All content is current and accurate
• No duplicate topics across documents
• Tested with 10 real prospect questions
• Knowledge gaps reviewed and addressed

What's Next

• Review your Chatbot Analytics to find knowledge gaps
• Check Chatbot Conversations to see what prospects are asking
• Connect your CRM to sync lead intelligence automatically

---

Questions? Contact us or email peter@parsley.id.