Difference between LangChain's PromptTemplate and ChatPromptTemplate

In LangChain, both PromptTemplate and ChatPromptTemplate are used to create prompts dynamically, but they are designed for different types of models.

Feature	PromptTemplate	ChatPromptTemplate
Purpose	Creates a single text prompt	Creates a structured chat conversation
Output	String	List of messages
Used With	Traditional LLMs	Chat models
Message Roles	No roles	Supports system, human ,user, system , assistant, ai , tool roles
Best For	Text completion models	Modern chat-based models

1. PromptTemplate

PromptTemplate generates a single formatted string.

Example

from langchain.prompts import PromptTemplate

prompt = PromptTemplate(
    template="Explain {topic} in simple terms.",
    input_variables=["topic"]
)

formatted_prompt = prompt.format(topic="Transformers")

print(formatted_prompt)

Output

Explain Transformers in simple terms.

The result is just a string.

How it is sent to an LLM

llm.invoke(
    "Explain Transformers in simple terms."
)

This style was common with older completion models.

2. ChatPromptTemplate

ChatPromptTemplate creates a conversation consisting of multiple messages.

Example

from langchain.prompts import ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are an expert teacher."),
    ("human", "Explain {topic} in simple terms.")
])

messages = prompt.format_messages(
    topic="Transformers"
)

print(messages)

Output

[
    SystemMessage(
        content="You are an expert teacher."
    ),
    HumanMessage(
        content="Explain Transformers in simple terms."
    )
]

Notice that the result is not a string.

It is a list of message objects.

How it is sent to a chat model

chat_model.invoke(messages)

Internally, the model receives:

System:
You are an expert teacher.

User:
Explain Transformers in simple terms.

Why ChatPromptTemplate Exists

Modern models such as:

• OpenAI GPT models
• Anthropic Claude models
• Google Gemini models

are chat-oriented.

They understand different message roles:

System: You are a helpful assistant.
User: What is Kubernetes?
Assistant: Kubernetes is...
User: Explain it simply.

PromptTemplate cannot naturally represent this structure.

ChatPromptTemplate can.

Side-by-Side Example

PromptTemplate

prompt = PromptTemplate(
    template="""
You are an expert teacher.

Question:
{question}
"""
)

prompt.format(
    question="What is Kubernetes?"
)

Produces:

You are an expert teacher.

Question:
What is Kubernetes?

ChatPromptTemplate

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are an expert teacher."),
    ("human", "{question}")
])

prompt.format_messages(
    question="What is Kubernetes?"
)

Produces:

[
    SystemMessage(
        content="You are an expert teacher."
    ),
    HumanMessage(
        content="What is Kubernetes?"
    )
]

Using Message Placeholders

One major advantage of ChatPromptTemplate is support for conversation history.

Example

from langchain.prompts import ChatPromptTemplate
from langchain.prompts import MessagesPlaceholder

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful assistant."),
    MessagesPlaceholder("chat_history"),
    ("human", "{question}")
])

Then:

prompt.format_messages(
    chat_history=[
        HumanMessage(content="What is Kubernetes?"),
        AIMessage(content="Kubernetes is a container orchestrator.")
    ],
    question="Who created it?"
)

Result:

System: You are a helpful assistant.
User: What is Kubernetes?
Assistant: Kubernetes is a container orchestrator.
User: Who created it?

This is essential for conversational applications.

When to Use Which?

Scenario	Recommended
Simple text generation	PromptTemplate
Chatbots	ChatPromptTemplate
RAG applications	ChatPromptTemplate
Agentic AI	ChatPromptTemplate
Conversation history	ChatPromptTemplate
Modern GPT/Claude/Gemini models	ChatPromptTemplate
Legacy completion models	PromptTemplate

Practical Recommendation

For most new LangChain projects in 2026:

ChatPromptTemplate

is the preferred choice because nearly all modern LLMs are chat-based, and it works naturally with:

• RAG pipelines
• Agents
• Tools
• Memory
• Multi-turn conversations
• System instructions

Use PromptTemplate mainly when you specifically need to generate a single text string or are working with APIs that expect plain text rather than chat messages.