Prompting Techniques: 9 Techniques for Better AI Results and Responses

Quick Overview

• Prompting techniques include zero-shot, one-shot, few-shot, chain-of-thought, self-consistency, least-to-most, role prompting, ReAct, and Tree of Thoughts.
• Zero-shot, one-shot, and few-shot prompting differ mainly in the number of examples given to the AI, from none to one or multiple examples.
• Reasoning-based techniques such as chain-of-thought, self-consistency, and least-to-most prompting help handle tasks that require multiple reasoning steps.
• Advanced techniques such as ReAct and Tree of Thoughts are useful for tool-based workflows, complex planning, and exploring multiple possible solutions.
• This guide covers how each prompting technique works, where to use it, practical examples, and how to choose the right technique based on task complexity and desired output.

Want to apply prompting techniques more effectively? Build practical skills with upGrad KnowledgeHut Generative AI and Prompt Engineering Course for Professionals. 

What are the main types of prompting techniques?

Prompting techniques are techniques of writing and structuring the instruction to an AI tool to make it understand the task and respond to it appropriately.

There are three broad groups of the key prompting techniques:

1. Foundational prompting techniques

Foundational techniques are the most basic approaches that can be used in everyday work with AI. The only difference between them lies in the number of examples shown to the AI before asking it to do something.

Examples of foundational prompting techniques are:

• Zero shot prompting
• One shot prompting
• Few shot prompting

2. Reasoning-based prompting techniques

Reasoning based techniques are helpful when a task involves several logical steps in order to achieve an appropriate outcome. They encourage AI models to consider the problem in a systematic manner.

Such types of techniques may be used in mathematics, analysis, problem solving, and similar types of tasks which require logical steps to arrive at an answer.

Examples of reasoning based techniques are:

• Chain of Thought prompting
• Self Consistency prompting
• Least to Most prompting

3. Advanced agentic prompting techniques

Advanced techniques can be applied in case of tasks related to planning, decision making, usage of tools or a variety of solutions. These types of strategies are especially important where AI programs have to reason through an issue instead of providing an answer directly.

Examples of such advanced techniques include:

• Role prompting
• ReAct prompting
• Tree of Thoughts prompting

Also Read: Prompt Engineering for Beginners: Step-by-Step Guide

 What are the main foundational prompting techniques? 

Foundational prompting is a basic way of instructing the AI. Foundational prompting is suitable for new users since it requires little prompt engineering and expertise.

There are three types of the foundational prompting technique which include zero shot, one shot and few shot prompting. The difference between the various types is the number of examples provided prior to performing the task.

Also Read: What is Prompt in AI?

1. Zero-Shot prompting 

Zero-shot prompting means instructing an AI to do something without giving any example of how it should be done. The word 'zero' in this context relates to the number of examples provided.

How it works: AI learns from massive data sets of text during the training process. Once a proper instruction is given, the model utilizes the learned knowledge to understand the context and respond accordingly.

Example: “Provide a summary of the benefits of working remotely in 100 words.”

The AI gets the instruction directly and finishes the task without needing any further instructions.

Best use cases of Zero-shot prompting

Zero-shot prompting works well for:

• General questions
• Summarization
• Brainstorming ideas
• Content generation
• Simple research tasks

2. One-Shot prompting

One-shot prompting consists of providing a single example before asking the AI to complete the task. The example serves as a reference point for the model to understand the desired format, tone or style.

Example:

Change the sentence to a more professional form.

Example: “Send me the report today.” → “Please share the report by the end of the day.” 

Now change: “I need the presentation quickly.”

The example above allows the AI to understand how to construct the answer.

One-shot prompting is useful when:

• You need a specific output format
• The task is relatively simple
• A single example clearly demonstrates the pattern

3. Few-Shot prompting

Few shot prompting provides the AI with various examples before making it perform a similar task. By observing various examples, the AI is able to learn the pattern and apply it to a new input. The various examples allow the AI to have a better idea of the way various reviews should be classified.

Among other foundational prompting techniques, few-shot prompting is considered the most reliable one.

Example:

Classify the reviews into Positive, Negative, and Neutral:

"The delivery was very fast." - Positive

"The product stopped working within one week." - Negative

"The product is good, but delivery was late." - Neutral

Now classify: "The quality is excellent, but the packaging is damaged."

Few-shot prompting is particularly useful when:

• Consistent formatting is required
• Accuracy is important
• Classification or extraction tasks are involved
• You need predictable outputs

Also Read: Prompt Engineering Best Practices for High Quality Outputs

What are the most effective reasoning-based prompting techniques?

Reasoning-based prompting techniques are designed for situations where a simple answer is not enough and the AI needs to analyze, reason, or solve a problem step by step. Some of these techniques are:

1. Chain-of-Thought (CoT) prompting

Chain of Thought or CoT prompting is a method where an AI model is asked to solve a problem through a series of reasoning processes.

Example:

What would be the total cost of 15 notebooks sold for ₹80 each with a 10% discount?

Calculation:

Original cost = 15 × ₹80 = ₹1,200

Discount = 10% of ₹1,200 = ₹120

Total cost = ₹1,200 - ₹120 = ₹1,080

CoT prompting can be helpful in solving problems that are mathematical in nature or logical questions or any questions that require a series of steps.

2. Self-consistency prompting

Self-consistency prompting involves using a problem-solving strategy in a slightly advanced way. Rather than having only one path of reasoning, the model can explore many paths and determine which answer is most consistent.

Example:

A shop gives a 20% discount on a product that costs ₹2,500. What is the final cost?

Two different calculation approaches may be considered here:

• Calculate the discount first and subtract it from the original price.
• Calculate the remaining 80% of the original price.

Both the approaches give ₹2,000.

Self consistency prompting is very effective in situations when several reasoning processes could be considered. It will be especially helpful in logical and mathematical questions.

3. Least-to-Most prompting

Least to Most involves the division of an apparently complex task into simple components. First, the AI solves the simplest components and subsequently uses the results to tackle the complex task.

Example:

Create a marketing strategy for a new online fitness program.

Rather than asking for the entire marketing strategy all at once, the task can be split as follows:

• Identify the target audience.
• Identify the audience's main problems.
• Select suitable marketing channels.
• Develop content ideas.
• Create a campaign structure.
• Define performance metrics.

Such divisions make a big task simpler to handle.

Least to Most prompting may be useful in research, planning, analysis, programming, and various other situations whereby one output depends on another earlier output.

Learn how prompting techniques fit into modern AI workflows with the Applied Agentic AI Certification Course [Best in 2026], designed for practical agentic AI applications. 

What advanced prompting techniques are used for complex AI tasks?

Advanced prompting is helpful for situations where an AI system requires additional information or different methods of reasoning, or interaction with other programs.

Below are some advanced prompting techniques.

1. Role prompting

The role prompting provides the AI with a certain role or perspective before being asked to perform some tasks. It allows directing on what type of the answer is required by providing the expected tone or expertise.

Example:

Act as an experienced SEO consultant. Provide five ways a small business can increase its website ranking. Use clear language and give examples for each suggestion.

In this case, the role, task and target audience for the AI is provided. This can help to receive a more relevant answer.

Role prompting can be helpful in content creation, teaching, brainstorming, business analysis, customer service, and explanations.

2. ReAct prompting

ReAct stands for Reasoning and Acting. This is an AI technique where the system combines reasoning and actions like performing a search query, accessing some information, or interacting with other tools.

The process here does not involve generating a response. Instead, the AI agent follows a process of:

Understanding the task → making a decision → taking an action → observing the outcome → and continue

Example:

Someone is asking an AI assistant the following question:

Identify three good hotels in Delhi for a business visit, compare them based on cost and location, and recommend the best one.

An AI system with a tool would have to perform search and gather information for comparing hotels before making the decision.

ReAct prompting is quite effective in tasks for AI agents where there is involvement of other tools or information.

3. Tree of Thoughts prompting

Tree of Thoughts prompting gives an opportunity to the AI system to discover several different approaches to a problem instead of being confined to a single approach.

Consider it as an attempt to create branches from some starting point. Every created branch means a possible solution that can be investigated and analyzed.

Example:

An organization needs to boost its sales online. Name three possible approaches, analyze their pros and cons, and provide recommendations on what would be the best approach for a small business having budget restrictions.

An AI may discover various options such as:

Boost online sales

• SEO strategy
• Paid advertising
• Email marketing

All these approaches may be analyzed according to such criteria as cost, efforts, expected results, etc.

Tree of Thoughts prompting can be helpful in strategic planning, decision-making, creative problem-solving, and other tasks that have multiple possible solutions.

Looking to apply AI beyond individual prompts? The AI-Powered Product Management Certification with Microsoft can help you use AI across practical product management workflows.

How do you choose the right prompting technique?

There is no single best technique for every situation. A practical approach is to start with the simplest method and add complexity only when the task requires it.

Based on task complexity

Task complexity is one of the simplest ways of determining the appropriate method. For a simple and familiar task, zero shot prompting may be enough.

For example: Summarize this article in five bullet points.

In case of the task, which requires a particular format or style, one-shot or few-shot prompting can serve as an example for the AI to follow.

For complex multistep tasks, reasoning-based approaches like Chain of Thought and Least to Most prompting might be appropriate. If the task requires using any tools or outside knowledge, then ReAct prompting is more suitable.

Task complexity	Suitable technique
Simple question or instruction	Zero shot prompting
Specific format or style	One shot prompting
Repeated pattern or classification	Few shot prompting
Multi step problem	Chain of Thought prompting
Problem with several smaller stages	Least to Most prompting
Task requiring a specific perspective	Role prompting
Task involving tools or actions	ReAct prompting
Multiple possible solutions	Tree of Thoughts prompting

Based on desired output quality

Another criterion in selecting a prompting technique would be the nature of the desired output quality.

1. Speed

For cases where speed is the main consideration, then one should use zero-shot prompting, which does not need much pre-training and takes seconds to generate useful output.

2. Accuracy

In cases where speed is not a concern but accuracy is, one could consider using few-shot prompting and self-consistency prompting techniques. More examples and reasoning can aid in minimizing mistakes.

3. Consistency

In cases where you require consistent answers in terms of the format, then the few-shot prompting technique is usually the most suitable. Examples help set up a pattern of answers for the AI.

4. Explainability

In some cases, the output answer may not be enough, it is also important to know the process behind an answer apart from knowing what the answer is.

Also Read: How to Use ChatGPT for Prompt engineering

Conclusion

The right prompting technique depends on the task, not on how advanced the technique sounds. Zero-shot prompting is often enough for simple requests, while one-shot and few-shot prompting help when the AI needs examples to follow.

For complex reasoning, techniques such as chain-of-thought, self-consistency, and least-to-most prompting can provide more structure. ReAct and Tree of Thoughts are better suited to tasks involving tools, planning, or multiple possible solutions. Start with the simplest approach and add more guidance only when the task requires it.

Have A Query? Get in Touch With Our Customer Support | upGrad KnowledgeHut