How to Build and Train a Custom AI Chatbot with FastBots.ai

Updated

FastBots.ai lets you create a chatbot trained on your own business content and deploy it to your website or messaging channels. The site describes a three-step setup flow and says you can be running in under 5 minutes. To follow this tutorial you need a FastBots.ai account (the site advertises a 14-day free trial with no credit card required) and the content you want the bot to answer from — website pages, uploaded files, or a ChatGPT-style source. The steps below cover setup, training, model choice, and deployment, plus what to check when something doesn't work.

Step 1: Create the bot and pick its knowledge sources

The first decision is where the bot's answers come from. FastBots.ai lists three knowledge source types:

  • Your own uploaded information — files and documents you provide directly.
  • ChatGPT — use a ChatGPT-style source to supply answers.
  • The Internet — pull in web content.

For most business sites, uploaded information plus your own website pages gives the most controlled results. If your answers live in PDFs, policy docs, or product sheets, upload those first; they tend to be more precise than crawled pages.

Expected result: a bot with at least one knowledge source attached. A bot with no sources has nothing to answer from, so don't skip this.

Step 2: Train on your website pages

FastBots.ai trains each chatbot on crawled pages of your website content. The published Essential plan allows 2,500 crawl pages per chatbot, so the crawl limit applies per bot, not across your whole account.

Practical approach:

  1. Point the crawler at the section of your site that actually answers customer questions — product pages, pricing, FAQ, shipping and returns — rather than the entire domain.
  2. After crawling, test the bot with real questions your customers ask and check whether the answers match the pages you intended.
  3. If answers are thin, add the missing pages as uploaded files instead of widening the crawl.

Verification: ask the bot a question whose answer exists only on one specific page. If it answers correctly, that page was crawled and indexed.

Step 3: Choose an AI model — and understand the credit cost

FastBots.ai lets you choose from models by OpenAI, Anthropic, and Google. Model choice directly affects how far your message credits go:

Model type Credit cost per reply
Standard AI models 1 credit
Advanced models 5–10 credits

On the Essential plan, 2,000 message credits per month are shared across all bots on the account. That means an advanced model at 10 credits per reply effectively gives you 200 replies, versus 2,000 with a standard model. Use standard models for high-volume FAQ handling and reserve advanced models for complex queries.

Step 4: Deploy to your website and messaging channels

FastBots.ai lists integrations with Zapier, Make, WhatsApp, Messenger, Instagram, WordPress, Slack, Telegram, and more — email is explicitly excluded. Two deployment paths:

  • Website: embed the bot on your site (WordPress is a listed integration) so visitors get answers in place.
  • Messaging: connect WhatsApp, Messenger, Instagram, Slack, or Telegram so the same trained bot answers where customers already are.

Verification: after deploying, send one test message per channel and confirm the reply matches what you saw in the builder. Channel-specific formatting is the most common place for surprises.

Optional: Let the bot take actions

The Zapier AI Actions feature lets bots call thousands of apps mid-conversation — checking orders, booking appointments, or sending emails. This is what turns a Q&A bot into a task-completing agent. Set it up only after the knowledge base answers reliably; an action bot with weak training will fail in ways that are harder to debug.

What the Essential plan includes

The published Essential plan is $39/month, or $33/month billed annually at $396 (saving $72), and includes:

  • 2 chatbots
  • 2,000 message credits/month, shared across bots
  • 2,500 crawl pages per chatbot
  • Knowledge sources, file uploads, reports, Zapier AI Actions, all LLMs
  • 3 team members

If you need more than 2 bots or more than 3 seats, check the pricing page for higher tiers before committing.

Common snags

  • Bot gives vague answers: the crawl likely missed the relevant pages, or the source content is ambiguous. Add a targeted upload.
  • Credits run out early: an advanced model is in use. Switch routine replies to a standard model.
  • Bot can't answer a question you know is on your site: confirm the page was within the crawl limit and isn't behind a login or script-rendered in a way the crawler can't read.
  • Integration connected but silent: test the channel directly; some messaging platforms require their own setup steps before the bot responds.