Creating an AI chatbot used to mean hiring a developer, wiring up a model, and maintaining it for months. Today you can create one that answers questions about your business in under 10 minutes, with no code.
iShort answer
The fastest way to create an AI chatbot is to train it on content you already have. Point SiteGPT at a website URL, sitemap, help center, or set of files. It uses retrieval-augmented generation (RAG) to answer from that content in your brand voice. Six steps, under 10 minutes, no code, and a 7-day free trial to test before you pay.
From a blank account to a chatbot embedded on your site, in under 10 minutes.
Create your account
Sign up for SiteGPT and start a new chatbot. Every plan has a 7-day free trial.
Add your content
Paste a website URL or sitemap, or add files, a help center, YouTube, or cloud storage (12+ sources).
Train the chatbot
Select the pages and sources, then let SiteGPT clean and index them.
Customize it
Set the name, welcome message, colors, logo, persona, and custom instructions.
Test it
Ask questions it should answer, then add anything missing and retrain.
Add it to your site
Paste the embed snippet or install the WordPress plugin.
6 steps
From signup to embedded on your site
<10 min
Typical time to launch
0
Lines of code required
12+
Content sources to train on
95+
Languages supported
7 days
Free trial on every plan
What is an AI chatbot?
A modern AI chatbot uses a large language model to hold a natural conversation, and it grounds every answer in your content with retrieval-augmented generation (RAG). A rule-based bot, by contrast, follows a script of predefined buttons.
The chatbot retrieves the most relevant passage from the material you trained it on, then writes an answer from it. The approach comes from Lewis et al. (2020), who found that retrieval-backed models produced more specific and factual language than a model answering from memory alone. That grounding is why the bot can go live in minutes and still answer accurately and on-brand.
Both answer visitors, but only one handles questions you did not script in advance.
Rule-based
- Follows predefined decision trees and buttons
- Only answers questions it was explicitly scripted for
- Breaks or dead-ends on unexpected phrasing
- Every new answer is a manual flow to build and maintain
AI / generative
- Holds a natural conversation with a language model
- Answers open-ended questions from your trained content
- Handles rephrasing and follow-ups without new rules
- Learns new answers when you add content and retrain
How do you create an AI chatbot step by step?
Train it on content you already have. These six steps use SiteGPT, and the whole flow takes under 10 minutes with no code.
Step 1: Create your account
Sign up for SiteGPT and start a new chatbot. Every plan includes a 7-day free trial, so you can build and test a chatbot before paying anything.
Step 2: Add your content
Choose what the chatbot should learn from. Paste a website URL and fetch its pages, submit a single page, or append /sitemap.xml to a domain to pull every indexed page at once.
You are not limited to a website. SiteGPT trains on 12+ source types, including YouTube videos, help centers (Zendesk, Freshdesk, Intercom, Confluence, Gitbook), files (PDF, DOCX, CSV, TXT, MD), and cloud storage like Google Drive, Notion, Dropbox, and SharePoint.
Step 3: Train the chatbot
Select the pages and sources you want to include, then start training. SiteGPT cleans and indexes the content so the chatbot can retrieve it during a conversation. That cleaning and indexing step is content ingestion, and how it copes with your messiest sources sets the ceiling on what the bot can answer.
Step 4: Customize it
Set the name, welcome message, and placeholder text. Adjust the colors and logo to match your brand, and write a persona plus custom instructions so replies sound like your company.
Step 5: Test it
Ask the chatbot questions it should be able to answer from your content. A good test set is your last 100 real customer questions. If an answer is thin or wrong, add the missing page or document and retrain.
Step 6: Add it to your site
Copy the embed snippet and paste it into your website, or install the WordPress plugin. The widget inherits the branding you set in Step 4, so it matches your site out of the box.
How do you create an AI chatbot with ChatGPT?
Connect the model to your own content. A raw ChatGPT model is a strong conversationalist, but it does not know your pricing, policies, or product details, and it can invent answers. How to stop a chatbot hallucinating covers the structural fixes for that.
RAG is how you make that connection, and SiteGPT runs it under the hood. It uses large language models for the conversation and grounds every reply in the content you trained it on. You get the fluency of ChatGPT with answers tied to your source material.
If you want to operate the chatbot from an assistant instead of a dashboard, SiteGPT also exposes a Model Context Protocol (MCP) connector. See how to build and run a chatbot from Claude for the walkthrough, or the remote MCP server page for the connection details.
Developers can skip the dashboard entirely and build the bot from the tools they already have open: Claude Code, Cursor, Codex, OpenClaw, and Hermes can each create and train a chatbot from a URL. SiteGPT for AI agents is the overview.
How do you train an AI chatbot on your own data?
Add your content as the source of every answer, in place of the model's general knowledge. With a RAG-based tool this is Steps 2 and 3 above: add the sources, select what to include, and the platform indexes it.
- Your data can include website pages, sitemaps, help center articles, files (PDF, DOCX, CSV, TXT, MD), YouTube videos, GitHub repos, and cloud storage like Google Drive, Notion, Dropbox, and SharePoint.
- No fine-tuning is required. Retrieval grounds the model in your content at answer time. It is faster, cheaper, and easier to update than training a custom model, and a 2024 Microsoft study found retrieval beat fine-tuning for adding knowledge (Ovadia et al.).
- On Growth and above, auto-sync re-scans sources on a schedule, so the bot tracks your content as it changes.
Can you create an AI chatbot in Python?
Yes. Developers can build one with a language-model API plus a retrieval layer, and that route makes sense when the chatbot is your product. For a support or FAQ bot it is usually the slow path. You own hosting, retrieval quality, widget code, and escalation logic for as long as the bot runs.
- The no-code route above ships the same RAG architecture in minutes rather than sprints.
- A middle path exists: drive a no-code bot from your own tooling. SiteGPT's remote MCP server lets Claude Code, Cursor, and Codex create, train, and update a bot from the terminal or CI.
- Reserve custom Python builds for chatbots that need product logic no platform exposes.
How do you add an AI chatbot to your website?
Copy and paste a snippet once the bot is trained. SiteGPT gives you a script snippet that works on any HTML page, plus a WordPress plugin.
The widget inherits the branding you set during setup, so it matches your site out of the box. Removing SiteGPT branding entirely is a paid add-on: $59/mo billed monthly or $39/mo billed yearly. Platform-specific setup (WordPress, Wix, Squarespace, Webflow) follows the same paste-the-snippet pattern.
If you are on WordPress, the install has its own steps and failure modes, covered in how to add a chatbot to WordPress. To pick the plugin first, the eleven options are ranked on plugin directory data in best chatbot for WordPress website.
How do you build a chatbot for a specific job?
Start from the general build above, then adjust it to the goal. If most of your volume is repeat questions, the AI FAQ chatbot tutorial shows the setup. For a service desk, follow the customer service chatbot tutorial, and for bookings, the appointment booking chatbot tutorial.
If the bot should capture and qualify leads, the funnel, qualification questions, and metrics are covered in chatbot for sales and marketing. For ready-made wording and flows, start from the chatbot templates.
What is a generative AI chatbot?
It is the technical name for the type of chatbot this guide creates. It generates each answer on the fly with a language model, rather than picking from a fixed list of scripted replies. Grounding that generation in your own content is what keeps the answers accurate.
How do you make a chatbot production-ready?
Add a human route, your chat tools, languages, and auto-sync.
- SiteGPT can escalate a conversation to a live agent on request, and your team gets notified when a visitor wants a person. How the main tools compare on context transfer and escalation triggers is covered in best AI chatbots with human handoff.
- On Growth and above, push conversations into Slack, Crisp, Messenger, Google Chat, Freshdesk Messaging, or Zoho SalesIQ. Zendesk handoff needs the Scale plan or above, plus Zendesk Suite Professional or higher.
- To answer on WhatsApp as well, add the WhatsApp add-on and connect a number. The steps are in how to create a WhatsApp chatbot.
- SiteGPT answers in 95+ languages from a single chatbot.
- On Growth and above, auto-sync re-scans your content on a schedule (monthly on Growth, weekly plus a daily scan on Scale) so answers stay fresh without manual retraining.
How much does it cost to create an AI chatbot?
You can build and test one free for 7 days on any plan (a credit card is required to start the trial). Paid plans start at $39/mo billed yearly, or $59/mo billed monthly:
| Plan | Billed monthly | Billed yearly (per month) | Chatbots | Messages/mo (up to) | Pages |
|---|---|---|---|---|---|
| Starter | $59/mo | $39/mo ($468/yr) | 1 | 4,000 | 1,000 |
| Growth | $129/mo | $79/mo ($948/yr) | 2 | 10,000 | 10,000 |
| Scale | $429/mo | $259/mo ($3,108/yr) | 3 | 40,000 | 50,000 |
Pick the plan by how many chatbots you run, how much traffic you get, and whether you need integrations.
| Scenario | Best pick | Why |
|---|---|---|
| Solo founder or one small site, just getting started | Starter | One chatbot, 4,000 messages, and manual retraining are enough to launch on a single site. |
| Growing team that needs Slack or Crisp and API access | Growth | Adds chat integrations like Slack and Crisp, API access, monthly auto-refresh, and up to 4 seats across 2 chatbots. |
| High-traffic site, several brands, or a Zendesk help desk | Scale | 40,000 messages, 50,000 pages, weekly plus daily auto-scan, webhooks, Zendesk handoff, and up to 10 seats. |
| Large org with high volume, many bots, or custom/security needs | Enterprise | Custom limits up to 500,000 pages, daily sync, priority support, and custom integrations. |
Yearly billing saves up to 40% against monthly (about 34% on Starter, 39% on Growth, and 40% on Scale). Removing SiteGPT branding is an add-on at $59/mo billed monthly or $39/mo billed yearly. See current details on the pricing page.
If your chatbot will handle protected health information (PHI), you need Enterprise. A signed HIPAA BAA is available on Enterprise only, and no PHI should reach the chatbot until it is executed and enablement is confirmed in writing.
A covered workspace also behaves differently from the build described above. Connected apps and chat integrations are off, PHI stays confined to end-user conversations, and the training content must be PHI-free.
Read the HIPAA docs before you plan the content sources. The full set of requirements is in what makes a chatbot HIPAA compliant. A healthcare bot that only answers from public information (service pages, scheduling guidance, insurance FAQs) needs no BAA and works on any plan.
Frequently asked questions
How do I create an AI chatbot for free? Start a 7-day free trial, train the chatbot on your website or documents, and test it in full before deciding on a paid plan. No code is required to build or test it.
How long does it take to create an AI chatbot? Usually under 10 minutes. Training on a website URL takes about as long as the crawl, and customization is a few clicks. There is no development cycle.
Can I create an AI chatbot without coding? Yes. A no-code tool like SiteGPT trains a chatbot on your content and gives you an embed snippet, so you never touch code.
How do I create an AI chatbot with ChatGPT? Connect a language model to your own content with RAG so answers stay accurate. SiteGPT does this for you: it uses large language models for the conversation and grounds replies in the content you trained it on.
What is the difference between an AI chatbot and a rule-based chatbot? A rule-based chatbot follows scripted paths and only answers what it was programmed to. An AI (generative) chatbot uses a language model plus retrieval over your content, so it can handle open-ended questions in natural language.
How do I add the chatbot to my website? Paste the embed snippet into your pages or install the WordPress plugin. The widget matches the branding you set during setup.
Sources
- Lewis, P., et al. (2020). Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks. NeurIPS 2020. The original RAG paper, cited for how retrieval grounds answers.
- Ovadia, O., et al. (2024). Fine-Tuning or Retrieval? Comparing Knowledge Injection in LLMs. EMNLP 2024. Supports the claim that retrieval outperforms fine-tuning for adding knowledge.
- SiteGPT pricing. Plan prices, limits, add-ons, and trial terms.
- SiteGPT HIPAA documentation. How a covered workspace behaves.