A chatbot is software that holds a text conversation with a customer. The original generation, still called rule-based, matches what the customer typed against patterns someone wrote and returns a pre-written reply. Newer chatbots generate the reply with a language model instead.

Chatbot is now an umbrella word covering two quite different technologies. A rule-based chatbot follows a decision tree, so it stays predictable and breaks on anything its author did not anticipate. Buyers often use the word for both kinds, which is worth checking before comparing two products that share the label.

How does a rule-based chatbot work?

A rule-based chatbot runs on patterns and branches. Someone defines the phrases a customer might use, maps each one to an intent, and writes the reply and the next question for each. The conversation then walks that tree.

The structure is why these systems behave so consistently. The same input produces the same output every time, the full set of possible replies can be read in advance, and nothing is generated at runtime that a person has not approved.

The same structure is why they break. A customer who phrases something unexpectedly, raises two issues at once, or changes subject mid-conversation leaves the tree, and the bot has only a fallback message to offer. Every additional case means another branch, so maintenance cost grows in step with coverage.

When is a rule-based chatbot still the right choice?

Predictability is worth paying for in narrow places. Where the script is fixed by regulation or policy, generating language at runtime adds risk without adding value. Rule-based still fits:

• Fixed disclosures that must be delivered word for word
  
  

• Simple routing where the goal is reaching the right team quickly
  
  

• Short transactional flows with a small, stable set of paths
  
  

• Environments with no approved route to a language model
  
  

The trap is using one for open-ended support. Coverage has to be built branch by branch while customer phrasing is effectively unbounded, so the backlog of unhandled cases grows faster than the team can close it.

Frequently asked questions

What is the difference between a chatbot and an AI chatbot?

A rule-based chatbot returns replies someone wrote in advance, matched by pattern. An AI chatbot generates the reply using a language model, working from a content source. The first is predictable and narrow, and the second handles unfamiliar phrasing and needs grounding to stay accurate.

Are rule-based chatbots obsolete?

Not obsolete, just narrower. They remain the better answer where every reply must be pre-approved, or where the flow is genuinely fixed. What has changed is the default: open-ended support used to mean building a decision tree, and now it rarely does.

Can you combine a rule-based chatbot with an AI agent?

Frequently, though it usually happens by accident rather than design. Many operations end up with a legacy tree handling routing and a newer system behind it, which works as long as one of them owns the handover. Where neither does, customers get dropped between them.