AI

What Is a Chatbot and How Does It Work? Rule-Based vs AI Assistant

November 15, 2026 · 8 min read

What is a chatbot and how does it work? Learn how rule-based bots differ from AI assistants, where each fails, and how to choose for your small business.

A chatbot is software that replies to customer messages in place of a person. There are two main kinds: a rule-based bot follows menus and keywords someone prepared in advance, while an AI assistant reads the meaning of a message and builds a reply from information you give it. Which one suits you depends on what your customers ask and how varied those questions are.

This guide explains how each type works, where each one gets stuck, and what to look at when you choose.

What a chatbot actually does

A chatbot sits at the end of a messaging channel, such as WhatsApp, Instagram or the chat box on your website, and receives incoming messages. It has three jobs:

  1. Work out what the customer wants.
  2. Find or produce a suitable answer.
  3. Send the answer back on the channel the customer used.

The difference shows up in step two. How the answer is found decides how useful the bot will be.

How a rule-based chatbot works

A rule-based bot works on "if this, then that". You draw a flow in advance: if someone writes "hello", send the menu; if they pick 1, send the opening hours; if the message contains the word "price", send the price list.

Strengths of rule-based bots

  • Predictable. What you wrote is what gets sent, so there is little risk of a surprise answer.
  • Simple to build. A small number of fixed questions can be set up quickly.
  • Good for fixed processes. "Press 1 to book, press 2 for opening hours" suits clear, stepwise tasks.

Weaknesses of rule-based bots

  • Wording trips them up. If a customer writes "how much is it, roughly?" instead of "price", the flow may not trigger.
  • They miss context. A message that contains the word "price" but is really a complaint can still receive the price list.
  • Maintenance is on you. Every new question or promotion means a new branch in the flow.
  • They trap people in the menu. A customer who asks something outside the menu can end up in an "I did not understand" loop.

How an AI assistant works

An AI assistant reads the meaning of a message instead of matching words. "How much is the red one?", "red price?" and "whats the cost for red" are understood as the same question. And a well-built assistant draws its answer from the information you provide, not from its own general knowledge.

In Blinkerbot the process looks roughly like this:

  1. It takes in your information. You let it read your website and upload documents such as a price list, policies and frequently asked questions.
  2. It reads the message. It understands text, voice notes, photos and PDFs.
  3. It answers from your information. On topics you have covered, it replies.
  4. It does not invent answers. If it has no answer, it says so and can hand the conversation to your team with a summary.

That last point matters. A general-purpose AI chat tool can state something it does not know in perfectly fluent language. In customer service that is unacceptable: if you promise a wrong price or an offer that does not exist, the promise is tied to you. The value of a business assistant lies less in how much it can say and more in knowing when to stay quiet. We cover this in more depth in what is an AI customer assistant.

The two approaches side by side

Take a message to a beauty salon: "Do you have a slot Friday evening for brows and lash lift, and is there parking?"

  • A rule-based bot catches the word "brows" or "appointment" and sends the main menu. The parking question is not in the menu, so it goes unanswered.
  • An AI assistant understands both services and the day, shows free slots based on how long those services take, and answers the parking question from the information you uploaded.

The reverse can also be true. When a customer only needs to make one short, fixed choice, such as "type YES to join the waiting list", a rule-based flow is enough and more predictable. So the two are tools that complement each other rather than rivals.

Which one should you choose?

Ask yourself these questions:

  • How varied are your messages? If the same five questions arrive every day, a menu may be enough. If every message is a little different, you need something that understands.
  • Is it only text? If customers send voice notes, photos or screenshots, a rule-based bot cannot interpret them.
  • Who will maintain it? If you do not have time to update branches every time something changes, an assistant that works from uploaded information creates less work.
  • How much risk can you tolerate? If you want no reply to be generated freely, fixed text may feel safer. But you can limit an assistant to your own information and make it hand over when it does not know.

For many businesses the best result is a mix: a ready-made flow for a specific campaign or fixed process, and an AI assistant for everything else. In Blinkerbot, the flow builder and the assistant work side by side in the same dashboard.

Common mistakes when setting up a chatbot

  • Hiding that it is a bot. There is no need. As long as the answers are fast and useful, customers rarely mind. Instead of hiding who is writing, show how to reach a person.
  • Leaving no way to reach a human. Showing the same menu to someone who typed "I want to speak to a person" loses customers.
  • Uploading incomplete or old information. The assistant is only as good as what you give it. When a price changes, update the document.
  • Automating everything. Complaints, refund requests and payment decisions are often better left to your team.
  • Ignoring channel rules. On WhatsApp, Instagram and Messenger, a reply window applies after the customer writes. Once it closes, outside the limited options Meta allows, you cannot reach out with free-form messages. A bot is there to answer quickly inside that window, not to get around it.

Does the customer care whether it is a bot?

Most customers care less about who is writing and more about whether the problem gets solved. A correct answer within a minute to a late-night "are you open tomorrow?" beats waiting until morning. But when a customer says they want a person, that has to be possible straight away. In a well-built system the assistant stops at that point and the conversation moves to your team.

What to watch after you launch

Setup is not the end. The first two or three weeks are the real test. In your inbox, look at:

  • Which questions did the assistant hand over? These are gaps in its information.
  • Does a customer ask the same thing twice? That suggests the first answer was not clear.
  • Who gets handed to a person? Complaints, special requests, or ordinary information the assistant simply did not have?

Read the numbers from your own reports. Do not use other businesses' results as your expectation.

How to set up a chatbot with Blinkerbot

  1. Create your account with the 14-day free trial. No card is needed.
  2. Connect your channels: WhatsApp, Instagram, Messenger, TikTok, Telegram or the chat on your website. Connections use official APIs.
  3. Teach the assistant your business: let it read your website and upload your price list, common questions and rules.
  4. Set the tone: write instructions for how formal to be, how long replies should be and which topics go to a person.
  5. If you want, draw a ready-made flow for a fixed process in the flow builder and leave the rest of the questions to the assistant.
  6. Write to it from your own phone as a customer would: ask a price, send a voice note, ask something it cannot know, say "I want to talk to a person". Fill the gaps, then go live.

For channel-specific setup, see the WhatsApp auto-reply guide and the Instagram DM automation guide. You can compare plans on the pricing page.

Frequently asked questions

Is a chatbot the same as an AI assistant?

Not exactly. "Chatbot" is the broad term, and both rule-based bots and AI assistants count as chatbots. A rule-based bot follows a prepared flow, while an AI assistant reads the meaning of the message and answers from your information.

Can an AI assistant give wrong information?

An assistant that answers only from the information you provide is set up to say it does not know instead of guessing. Even so, read its conversations in the first weeks and fill in what is missing.

Does a small business need a chatbot?

If you keep answering the same questions or miss messages outside working hours, it can help. If you receive a few messages a day, replying by hand may be perfectly reasonable.

Do I need to know how to code?

No. In Blinkerbot you set things up by uploading your information, writing instructions and connecting your channels. No code is needed.

What happens if a customer wants to talk to a person?

The assistant hands the conversation to your team, and you or a colleague can take it over from the shared inbox with one tap. Once someone has taken over, the assistant stays quiet in that conversation.

Blinkerbot

Get messages answered while you're away.

Let an AI assistant answer your WhatsApp, Instagram and website messages, and hand over what it doesn't know. 14 days free, no credit card.