AI Agent vs Chatbot vs AI Assistant: What's the Difference?

A chatbot holds a conversation, an AI assistant helps by answering each of your requests, and an AI agent receives a goal and then acts on its own, step by step, using tools. So the real difference comes down to two questions: who decides the next step, and can the software act beyond the conversation?

Three words, three levels of autonomy

These terms are often used interchangeably, including by the companies that make the products. Yet they describe different things, which you can place on a scale of autonomy.

The chatbot: a conversation

A chatbot is a program that talks with you in writing or by voice. The idea is old: ELIZA, created by Joseph Weizenbaum at MIT in the mid-1960s, already imitated a psychotherapist by rephrasing what the user said.

Many of the chatbots you meet are still tightly scoped: the chat window on a retail site, your phone carrier’s customer service, the bot that asks for your order number. They often follow pre-written scripts or answer from a list of frequently asked questions. When your request falls outside that scope, they hand you over to a human.

The AI assistant: all-round help

An AI assistant is a far more capable chatbot, built on a large language model. It understands a huge range of requests: it writes, summarizes, translates, explains and analyzes documents you give it. ChatGPT, Claude, Gemini and Le Chat are assistants.

Its built-in limit: it answers, then waits. You lead the conversation, you decide the next question, you copy the output to wherever it needs to go. Even when it searches the web to answer you, you’re the one steering.

The AI agent: a mission

An AI agent receives a goal, not a question. It breaks the work into steps, chooses which tools to use (browser, spreadsheet, email, code execution), acts, observes the result and repeats until the task is done. Under the hood, it relies on tool calling: the model writes a structured request that a program executes on its behalf.

To understand this loop and its limits in detail, our full guide, what is an AI agent? Definition, how it works and real examples, starts from the beginning.

The differences at a glance

Criterion	Chatbot	AI assistant	AI agent
What you give it	A question in a specific area	One request, then another	A goal to reach
Who decides the next step	A pre-written script	You, with every message	The agent itself
Can it act outside the chat	Rarely	A little: web search, file creation	Yes: browse, click, fill in, send
Steps per request	One	One, sometimes a few	Many, sometimes dozens
Typical example	Order tracking on a retail site	Rewording an email	Comparing offers and building a table
Main risk	An off-topic answer	False information (hallucination)	A wrong action with real consequences
Your role	Ask the question	Steer the conversation	Set the goal and approve

In real life, the lines blur

As of October 22, 2026, most major assistants offer an agent mode, according to their makers’ websites. The same product can therefore be an assistant when you ask it to fix a paragraph, and an agent when you ask it to research, compare and compile information across several sites.

The table below shows the features declared by a few vendors, including whether an agent mode is available:

Assistant	Free	Paid from	Web search	Agent mode	Voice	Images
ChatGPT	Limited	$8/mo	Yes	Yes	Yes	Yes
Claude	Limited	$20/mo	Yes	Yes	—	—
Microsoft Copilot	Limited	$9.99/mo	Yes	—	Yes	Yes
Alexa+	Limited	$19.99/mo	Yes	Yes	Yes	—
Manus	Limited	$20/mo	Yes	Yes	—	—

A few cases to help you find your way:

• Voice assistants were long just command runners (“set a timer”). The new generations, such as Alexa+, run on language models and list an agent mode, according to Amazon. We’re moving from the classic voice assistant toward something closer to an agent.
• Agentic browsers include an agent that clicks and fills in forms for you. We cover them in AI browser agents: when AI uses the web for you.
• General-purpose agents, such as Manus, are built first and foremost to carry out complete tasks rather than to chat.

Marketing adds to the confusion: the word “agentic” is sometimes slapped on products that actually follow a fixed script. We untangle the jargon in agentic AI: what the buzzword really means.

Check yourself A delivery company's website bot asks for your order number, then shows where your parcel is. Chatbot, assistant or agent?

It’s a chatbot. It follows a set script within a narrow scope and simply looks up a piece of information to show you. It doesn’t choose its own steps and doesn’t act outside that predefined path.

Three questions to spot a real agent

When a new tool is pitched as an “agent”, ask yourself these three simple questions. In most cases, they’re enough.

1. Am I giving it a goal or a question? If you have to prompt the tool again at every step, it’s an assistant. If you describe an end result and it organizes itself to get there, you’re dealing with an agent.
2. Can it act outside the chat window? Opening a website, filling in a form, creating a file in your cloud storage, sending a message: these are actions. A tool that only produces text is still an assistant, however capable.
3. Does it change plans when something fails? A fixed script stops or shows an error. An agent tries another route, for instance another source when a page doesn’t respond.

If the answer to all three is yes, the tool does behave like an agent. That says nothing about its reliability: an agent can take a wrong step, misread a page or make up information. The more it acts, the more human oversight matters.

Verdict: which one to use, and for whom?

There’s no “best” tool in absolute terms: it all depends on what you want to do and how much control you want to keep.

A chatbot is right if you have a simple question in a specific context: tracking an order, booking an appointment, finding information about a service. There’s nothing to install; the website provides it.

An AI assistant is right if you want to write, understand, summarize, translate or think out loud. It’s the best starting point for beginners: you stay in control at every step and see immediately what the AI produces. To pick yours, our AI assistant comparison tool lines up free tiers, prices and features side by side.

An AI agent is right if you have a multi-step, repetitive or lengthy task: gathering information from several sites, preparing a comparison table, sorting documents. It saves time, as long as you check the result and keep approval over sensitive actions. Before delegating, our Can I trust an agent with this? tool helps you gauge the risk of a specific task.