FAQBotby Gator Labs

Uploading documents

Strong answers come from strong inputs. Start by uploading the content you already rely on to explain things to customers and new team members.

What to upload first

FAQs & help content

General FAQs, onboarding guides, “how do I…?” questions, and any support responses you send repeatedly.

Pricing & services

Pricing pages, menus, plan features, scope of work, and detailed breakdowns of what’s included or excluded.

Policies & rules

Refunds, cancellations, shipping, lead times, booking rules, and anything where a clear explanation avoids extra back-and-forth.

Example of a clean upload set

A simple, well-organised upload set works far better than one giant exported file. Here‘s a real structure that consistently produces excellent results:

Example for SaaS, service businesses or small teams

Brand – Getting started guide.md
Brand – Pricing & plans.md
Brand – FAQs (general).md
Brand – Support policy & response times.md
Brand – Cancellations & refunds policy.pdf

Keep each document focused on one topic, and use clear headings for natural chunking.

For local businesses (bakeries, salons, cafés), replace these with: menus, custom order rules,pricing ranges, lead times, and frequently-asked booking questions.

Supported formats & best practices

Supported formats

  • PDF files (with selectable text)
  • Plain text (.txt)
  • Markdown (.md)
  • Word/Google Docs (export as PDF)

Best practices

  • Split long documents into topics (pricing, onboarding, features, policies).
  • Keep each file focused on a single area or theme.
  • Avoid image-only PDFs or scans with unreadable text.
  • Remove or clearly archive outdated content to avoid conflicts.

What happens after upload

Once uploaded, FAQBot breaks your content into sections, turns those into embeddings, and stores them in a private index. When someone asks a question, the system searches those embeddings, selects the most relevant pieces, and builds an answer grounded in your content — complete with citations.

If answers look wrong, start by checking:
– the processed document in Documents
– the citations shown in the chat console.