CRISPE Prompt Framework: Explore Options and Variants

Study guide

CRISPE Prompt Framework: Explore Options and Variants

CRISPE is a prompt framework created by Matt Nigh and published as a GitHub README on 7 February 2023. Its five letters stand for Capacity and Role, Insight, Statement, Personality and Experiment. This course follows the original version, because many websites expand the acronym differently and the same letters can mean slightly different things depending on where you read them. The framework works as a build order: role first, then context, then the task, then tone, then a request for options. Skipping early steps makes the AI guess.

This course is for busy professionals who paste prompts into ChatGPT, Claude or Gemini and want several options to compare instead of one obvious answer. You will learn how to set a sharp role, paste real customer quotes and research into the Insight slot, state your request plainly, set a personality that matches your brand, and ask for labelled variants you can mix into one final version. Each lesson covers the why, the how, a worked example in a real job, and a common mistake. By the end you will know when CRISPE is the right tool and when a leaner framework will get you there faster.

What is a prompt framework?

A prompt is the request you type into an AI tool such as ChatGPT, Claude or Gemini. A prompt framework is a reusable structure for that request: a short checklist of the parts a good prompt contains, usually named with an acronym so it is easy to remember. Why it helps: the AI fills in whatever you leave out with generic guesses. A framework makes sure you include the parts that change the answer, such as who it is for, what you want to achieve and the format you need.

Frameworks make prompts repeatable: you fill the same parts each time, save the prompt as a template and share it with colleagues. They are not magic words: the value comes from the detail you put in each part. They help most when a request is complex or open to interpretation; a quick, simple question can stay short. There are well over sixty named frameworks, and most of them remix the same ingredients: role, task, context, audience, format, constraints and examples.

In this course you learn CRISPE, which is a fill-in-the-slots framework: an acronym where each letter stands for one part of your prompt. It is best for: exploratory and creative work where you want several variants to compare.

Where CRISPE comes from

CRISPE was created by Matt Nigh and published as a GitHub README on 7 February 2023, which places it among the first prompt frameworks ever shared publicly. The five letters stand for Capacity and Role, Insight, Statement, Personality and Experiment, and they are meant to be used in that order. Many websites expand the letters differently, so the same acronym can mean slightly different things depending on where you read it. This course uses the original version from Matt Nigh.

The framework works as a build order: role first, then context, then the task, then tone, then a request for options. Skipping early steps makes the AI guess. If you leave out the role, the model writes as a general assistant. If you leave out the insight, it invents your market. If you leave out the experiment step, you get one answer and no way to compare.

The final letter, Experiment, is what separates CRISPE from a plain instruction. It asks the AI for several variants so you can choose rather than accept the first answer. That single habit, asking for options, is the most portable idea in the whole framework. You can add it to any other prompt you write.

Capacity and Role: set the expert and the hat

Capacity and Role are the first two letters of CRISPE and they work as a pair. The Role names the expert, the Capacity describes what that expert can do. A strong Role line is specific. For example: act as a senior copywriter with ten years in B2B software. The field, the seniority and the audience all shape the answer. Capacity adds knowledge, tools and standards. Knowledge is what the expert knows, tools are what they use, standards are what they check their work against.

A sharp role is what lets later variants differ in a meaningful way. Swap a precise role and you get a genuinely different angle. Swap a vague one and the answers blur together. This is why the role belongs before the task in your prompt. The AI then reads the job as a specialist rather than a general assistant.

The most common mistake is naming a role with no capacity. Act as an expert gives the AI nothing to work with, so it falls back on generic output. A better line names the field, the years and the audience, and says what the expert checks. For example: act as a customer support lead with eight years in subscription software, who writes replies that are short, calm and free of jargon.

Insight: give the AI your facts, not guesses

Insight is the third slot in CRISPE and holds the background: who the customer is, what they believe, and what competitors say. It is the difference between the AI guessing your market and the AI working from your market. Specificity is the rule. One real customer quote beats three paragraphs of description, because a quote carries the customer's own words while a summary carries your assumptions.

This is the slot where you paste research: reviews, interview notes, survey comments, support tickets, competitor pages. Paste the raw material rather than describing it, so the details survive. A weak Insight slot reads like a conclusion, for example customers value ease of use. A strong one reads like evidence, for example a quote about setup taking a week. The AI can build on evidence, not on verdicts.

Competitor claims belong in Insight too. If a rival says live in five minutes, the AI needs that claim in front of it to draft positioning that answers it. Insight works with the other slots. Capacity and Role set who the AI is, Insight sets what it knows, Statement sets the task, Personality sets the tone, Experiment asks for variants.

Statement and Personality: state the task, set the voice

Statement is the actual request, stated plainly. It should be one clear line, not a hint buried in background. If you want five post ideas, write that. A weak Statement makes the model guess the task. For example, instead of saying I need something for the launch, write: give me five LinkedIn post openings for the launch of our new reporting feature.

Personality is the voice of the answer: witty, calm, bold or academic. It changes how the answer sounds, not the facts it contains. Keep the personality consistent with your brand or your audience. A short tone line reused across requests keeps your content feeling like one voice. Personality works best as a few clear words, such as witty and warm, short sentences, no hard selling. Long paragraphs about brand values dilute the instruction.

S and P work as a pair. The Statement says what to do, the Personality says how it should sound when it comes back. A missing Personality line gives flat, generic answers you have to rewrite by hand. A clear one saves you an editing pass.

Experiment: ask for labelled variants

Experiment is the E in CRISPE and it means asking for several versions or approaches in a single prompt, for example, give me five options, each with a different angle. Asking for variants lets you compare options instead of accepting the first answer the AI produces, which is usually just the most obvious one.

Always ask the model to label each variant with its angle, such as playful, serious or customer story, so you can tell them apart and mix the best parts. Labelled variants turn a flat answer into a menu of choices, which is far easier to review and defend in a real work setting. A common mistake is asking for five options without naming the angles, which produces five paragraphs that all sound the same.

Once you have labelled variants, the real skill is mixing: take the strongest line from each angle and combine them into one final version. For example, you might keep the opening from the customer story angle, the call to action from the serious angle, and the rhythm from the playful one. That final version is often stronger than any single variant.

CRISPE in your job: when to use it and when to skip it

CRISPE is best when you do not know what you want yet. The six parts open up thinking and give you options. If you already know the output you want, a leaner framework like CO-STAR will get you there faster. The main weakness of CRISPE is weight. Six parts for a quick task is too much. A short question or a fast rewrite does not need a full CRISPE prompt. Match the framework to the size of the task.

The Experiment step is the most portable habit in the whole course. Asking for three versions costs nothing, one extra line in your prompt. You can add it to CO-STAR, to a plain question, or to any other framework. That single line often improves the result more than any other change.

A worked example helps you see the parts in action. A teacher prompt used Capacity and Role, Insight, Statement, Personality, and Experiment. Each part did a job, and the result was three ready-to-use lesson openings. The anatomy of a CRISPE prompt is a checklist. Capacity and Role set who the AI is. Insight gives context. Statement says what you want. Personality sets tone. Experiment asks for options. Walk through each slot before you send. Choosing the right framework is a skill. CRISPE for open, uncertain tasks. CO-STAR for clear, fast ones. The Experiment line for both. That decision saves you time and gets better answers.

Frequently asked questions