Prompt Engineering Techniques

These techniques are not recommended for reasoning models like gpt-5 and o-series models.

Learning how to write the best AI prompts can be difficult. Prompt construction is more of an art than a science, often requiring experience and intuition to craft successful AI prompts. The goal of this article is to help you get started with this learning process. It attempts to capture general concepts and patterns that apply to all GPT models. However, it is important to understand that each model behaves differently, so the learnings might not apply equally to all models.

---

Basics

This section covers the basic concepts and elements of GPT prompts.

Text prompts are how users interact with GPT models. As with all generative language models, GPT models attempt to produce the next series of words that are most likely to follow from the previous text. It is as if we are saying, "What is the first thing that comes to your mind when I say <prompt>?" The examples below demonstrate this behavior. Given the first words of famous content, the model is able to accurately continue the text.

As you develop more complex prompts, it is helpful to keep this fundamental behavior in mind. Regardless of the prompt that is provided, the model is simply responding with what it determines is most likely (given its training data and training targets). If you ask a question in your prompt, the model is not following a separate "Q&A" code path. It appears to answer the question because an answer is the most likely response for the given question as input.

---

Prompt components

When using the Completion API, while there is no differentiation between different parts of the prompt, it can still be useful for learning and discussion to identify underlying prompt components. With the Chat Completion API, there are distinct sections of the prompt that are sent to the API in the form of an array of dictionaries with associated roles: system, user, and assistant. This guidance focuses more generally on how to think about prompt construction rather than providing prescriptive guidance that is specific to one API over another.

It is also important to understand that while there could be other valid ways to dissect prompts, the goal of this breakdown is to provide a relatively simple way to think about prompt construction. With the Completion API, all the components are optional, but at least one must be present and most prompts include more than one component. There can be some gray area between components as well. The order presented below roughly corresponds to how commonly each component is used, from most to least.

Instructions

Instructions are likely the most commonly used prompt component. Instructions are simply that: instructions to the model on what to do. While simple in concept, they can be complex in practice. Constraint-heavy tasks like Regex Generator show how small wording changes can reshape outputs. The following table illustrates this by showing both simple and more complex instructions, in two different forms.

Primary content

Primary content refers to some sort of text that is being processed or transformed by the model. Primary content is typically used in conjunction with instructions. A simple example would be language translation. In the example below, the English text would be considered the primary content while "Translate to French:" is an instruction.

Primary content can also be much longer. In the following example, the primary content is the introduction section from the Wikipedia entry on time complexity, which is almost 300 words long. This content is abbreviated in our table for ease of display.

GPT models can also handle primary content that is structured. In the example below, a TSV about beers in Belgium (abbreviated for display in this article) is passed to the model as part of the prompt. It is capable of properly interpreting the content and answering questions about the data. If you want a fast structured-data test, SQL Terminal is a simple way to practice schema-specific prompts.

Examples

Successful prompts often rely on the practice of one-shot or few-shot learning. This refers to the inclusion of one or more examples of the desired behavior of the model, typically by including input and output pairs. This is not learning in the sense that the model is permanently changed, but rather that the examples better condition the model to respond as desired for only the current inference. The use of prompts with no examples is sometimes referred to as zero-shot learning. Note that with the Chat Completion API few-shot learning examples are typically added to the messages array in the form of example user and assistant interactions after the initial system message.

The example above illustrates the utility of few-shot learning. Without the examples, the model seems to be guessing at the desired behavior, while the examples cleanly show the model how to operate. This also demonstrates the power of the model, it can infer the category of label that is wanted, even without a "basketball" label in the examples.

Cue

Cues act as the "jumpstart" for the output of the model, helping to direct the model to the desired output. It is often a prefix that the model can build onto. In the example below, we show the cue directing the story to include key elements. Cues are often used along with instructions, but not always. In the examples below, the model is being used to summarize a long announcement email. A cue is used in one case to help focus the output of the model and to suggest a specific output format (bullet points) in the other.

In the context of the Chat Completion API, instructions take the form of the system message, and few-shot examples are indicated by providing a series of example user and assistant interactions to help prime the model.

Supporting content

Supporting content is information that the model can utilize to influence the output in some way. It differs from primary content in that it is not the main target of task, but it is typically used along with primary content. Common examples include contextual information such as the current date, the name of the user, user preferences, and so on.

The example below uses supporting content to help organize a set of planned workshops for the user. Without the supporting content (important topics), the model simply lists off the workshops (truncated for display). When informed of my important topics, the model is able to accurately group the meetings.

---

Scenario-specific guidance

The techniques in this section will teach you strategies for increasing the accuracy and grounding of responses you generate with a large language model (LLM). It is, however, important to remember that even when using prompt engineering effectively you still need to validate the responses the models generate. Just because a carefully crafted prompt worked well for a particular scenario does not necessarily mean it will generalize more broadly to certain use cases. Understanding the limitations of LLMs is just as important as understanding how to leverage their strengths.

This guide does not go in-depth into the mechanics behind the message structure for Chat Completions. If you are not familiar with interacting with Chat Completions models programmatically, we recommend reading a how-to guide on the Chat Completion API first.

Note: All of the examples in this section of the guide were tested against a base GPT-5 model in English. Some techniques may produce different results with newer models like gpt-4o and gpt-4.5. If you are reading a localized version of this article in another language, these responses represent a localized translation of the English results. To learn more about potential limitations depending on what language you are using to prompt a model, consult the Responsible AI transparency note.

Few-shot learning

A common way to adapt language models to new tasks is to use few-shot learning. In few-shot learning, a set of training examples is provided as part of the prompt to give additional context to the model.

When using the Chat Completions API, a series of messages between the user and assistant (written in the new prompt format), can serve as examples for few-shot learning. These examples can be used to prime the model to respond in a certain way, emulate particular behaviors, and seed answers to common questions.

The table above describes the basic prompt structure, but for a detailed explanation of the exact prompt format you will need to refer to the Chat Completion guide.

Non-chat scenarios

While the Chat Completion API is optimized to work with multi-turn conversations, it can also be used for non-chat scenarios. If you want a minimal, execution-style check, Python Interpreter is a good way to test instruction precision. For example, for a sentiment analysis scenario, you might use the following prompt:

System message
You're an assistant designed to analyze sentiment from speech data. Users will paste in a string of text and you'll respond with an assessment of the response of the speaker. Rate on a scale of 1-10 (10 being highest). Explain why this rating was given.

User
hi thank you for calling contoso pharmacy who am i speaking with today i understand your frustration mary um can you tell me what exactly you're trying to accomplish OK uh certainly happy to check that for you one moment please i see here that you were on a generic form of lipitor is that right OK uh so i see that your doctor stopped prescribing these drugs in twenty twenty one actually OK uh yeah i'd be happy to check that for you uh because sometimes there's a gap in the system and it just doesn't get reported so let me take a look here just a moment so i'm seeing here that your doctor had these drugs prescribed to you from twenty twelve through twenty twenty one OK um yeah according to the latest records uh provided by doctor's office you're now on a different drug metformin would you like us to go ahead and fill that prescription for you for pick up a bit of form in five hundred milligrams uh you're very welcome mary please let us know if there's anything else we can do for you today.

Assistant
Sentiment rating: 8
The speaker in this text seems to be polite, helpful, and patient while addressing the concerns of the customer. They take the time to check the information, provide clear explanations, and offer assistance. However, the rating is not a 10 because there is still a hint of uncertainty in some parts of the conversation, which might not convey complete confidence to the customer.

Start with clear instructions

The sequence information appears in the prompt matters. This is because GPT style models are built in a certain way that defines how they process the input. Research suggests that telling the model the task you want it to do at the beginning of the prompt, before sharing additional contextual information or examples, can help produce higher quality outputs.

Note: Although following this technique is still generally recommended, testing showed that the model response with ChatGPT and GPT-5 models was the same regardless of whether the technique is used. In the example below, adding the statement "Several sources ... eruption" at the beginning or the end of the prompt did not cause any change in the final model response.

Repeat instructions at the end

Models can be susceptible to recency bias, which in this context means that information at the end of the prompt might have more significant influence over the output than information at the beginning of the prompt. Therefore, it is worth experimenting with repeating the instructions at the end of the prompt and evaluating the impact on the generated response.

Prime the output

This refers to including a few words or phrases at the end of the prompt to obtain a model response that follows the desired form. For example, using a cue such as "Here is a bulleted list of key points:\n- " can help make sure the output is formatted as a list of bullet points.

In the above prompt, the text "One possible search query is:" primes the model to produce a single output. Without this cue the model produces several search queries as output.

Add clear syntax

Using clear syntax for your prompt, including punctuation, headings, and section markers, helps communicate intent and often makes outputs easier to parse.

In the example below, separators (--- in this case) have been added between different sources of information or steps. This allows the use of --- as a stopping condition for generation. In addition, section headings or special variables are presented in uppercase to differentiate them.

If you are not sure what syntax to use, consider using Markdown or XML. The models have been trained on a large quantity of web content in XML and Markdown, which might provide better results.

Break the task down

Large language models often perform better if the task is broken down into smaller steps. For example, in the search query prompt referenced earlier, the prompt can be restructured so that the model is first instructed to extract relevant facts, and then instructed to generate search queries that can be used to verify those facts.

Use of affordances

Sometimes we can get the model to use an affordance instead of relying on its own parameters for information and answers. Search, for example, can be an affordance to help mitigate against fabricated answers, and to get up-to-date information.

One simple way to use an affordance is to stop generation once the affordance calls are generated by the model, then paste the outcomes back into the prompt. Below is an example of what the follow-up call could be after doing the SEARCH calls above. Notice how we pasted the search results into the prompt and replaced the previous SEARCH calls.

System message
You're an AI assistant that helps people find information. You'll read a paragraph, extract factual claims, and then use search engine results to fact-check them.
---
PARAGRAPH
John Smith is married to Lucy Smith. They have five kids, and he works as a software engineer at Microsoft. What search queries should I do to fact-check this?
---
FACTUAL CLAIMS
- John Smith is married to Lucy Smith
- John and Lucy have five kids
- John works as a software engineer at Microsoft
---
Here are various search queries issued to research the claims above:
SEARCH QUERIES
- John Smith married to Lucy Smith
- John Smith number of children
- John Smith software engineer Microsoft
---
Here are snippets of the search results:
SNIPPETS:
[1] ... John Smith's wedding was on September 25, 2012 ...
[2] ... John Smith was accompanied by his wife Lucy to a party
[3] John was accompanied to the soccer game by his two daughters and three sons
[4] ... After spending 10 years at Microsoft, Smith founded his own startup, Tailspin Toys
[5] John M is the town smith, and he married Fiona. They have a daughter named Lucy
---
Given the snippets, fact check each of the factual claims above:

Chain of thought prompting

Important: This technique is only applicable to non-reasoning models. Attempting to extract model reasoning through methods other than the reasoning summary parameter are not supported, may violate acceptable use policies, and may result in throttling or suspension when detected.

This is a variation on the break the task down technique. Instead of splitting a task into smaller steps, in this approach, the model response is instructed to proceed step-by-step and present all the steps involved. Doing so reduces the possibility of inaccuracy of outcomes and makes assessing the model response easier.

Specifying the output structure

Using your prompt to specify the structure of the output can have a significant effect on the nature and quality of the results. Sometimes system message inputs like "only write true facts" or "do not fabricate information" might not be adequate mitigations. Instead, asking for the model response to also include citations can help reduce the prevalence of incorrect responses.

If you instruct the model to cite the source material when it makes statements, those statements are much more likely to be grounded. Asking for citations makes it so that the model must make two errors every time it generates a response: the first error is the fabricated response, and the second is the bad citation. The closer the citation is to the text it supports, the shorter the distance ahead the model needs to anticipate the citation, which suggests that inline citations are better mitigations for false content generation than citations at the end of the content.

Similarly, if you ask the model to extract factual statements from a paragraph, it might extract compound statements such as "X is doing Y AND Z" (which can be harder to verify). This can be avoided by specifying an output structure such as (entity1, relationship, entity2).

The examples below show the use of citations and directing the model response to fit a defined structure.

Temperature and top_p parameters

Changing the temperature parameter changes the output of the model. The temperature parameter can be set between 0 and 2. A higher value, for example 0.7, makes the output more random and produce more divergent responses, while a lower value, like 0.2, makes the output more focused and concrete. A fictional story could be generated using a higher temperature, whereas to generate a legal document it is recommended to use a much lower temperature. Top probability is another parameter and is similar to temperature in that it also controls the randomness of the model response, but it does so in a different way. The general recommendation is to alter one of these two parameters at a time, not both.

Provide grounding context

One of the most effective ways to provide reliable answers is to give the model data to draw its responses from (grounding data). If your use case relies on up-to-date, reliable information and is not purely a creative scenario, it is strongly recommended to provide grounding data. In general, the closer you can get your source material to the final form of the answer you want, the less work the model needs to do, which means there is less opportunity for error.

---

Best practices

• Be specific. Leave as little to interpretation as possible. Restrict the operational space.
• Be descriptive. Use analogies.
• Double down. Sometimes you might need to repeat yourself to the model. Give instructions before and after your primary content, use an instruction and a cue, and so on.
• Order matters. The order in which you present information to the model might impact the output. Whether you put instructions before your content ("summarize the following...") or after ("summarize the above...") can make a difference in output. Even the order of few-shot examples can matter. This is referred to as recency bias.
• Give the model an "out." It can sometimes be helpful to give the model an alternative path if it is unable to complete the assigned task. For example, when asking a question over a piece of text you might include something like "respond with 'not found' if the answer is not present." This can help the model avoid generating false responses.

---

Space efficiency

While the input size increases with each new generation of GPT models, there will continue to be scenarios that provide more data than the model can handle. GPT models break words into tokens. While common multi-syllable words are often a single token, less common words are broken into syllables. Tokens can sometimes be counter-intuitive, as shown by the example below which demonstrates token boundaries for different date formats. In this case, spelling out the entire month is more space efficient than a fully numeric date.

Given this limited space, it is important to use it as efficiently as possible.

• Tables: As shown in the examples in the previous section, GPT models can understand tabular formatted data quite easily. This can be a space efficient way to include data, rather than preceding every field with a name (such as with JSON).
• White space: Consecutive whitespaces are treated as separate tokens, which can be an easy way to waste space. Spaces preceding a word, on the other hand, are typically treated as part of the same token as the word. Carefully watch your usage of whitespace and do not use punctuation when a space alone will do.