The RISE Framework
Role, Input, Steps, Expectation — four pieces that turn a request into a procedure. RISE is the framework to reach for when you want the AI to follow a defined process, whether that is walking you through a physics problem or building an assignment one stage at a time.
What is the RISE framework?
The RISE framework is a recipe for prompts that ask the AI to do something in order rather than just answer a question. Instead of describing an outcome and hoping the model takes a sensible path to it, you give it four things: the Role it should play, the Input it has to work from, the Steps it should take and in what sequence, and the Expectation that describes what a finished result looks like.
Large language models like ChatGPT, Claude, and Gemini generate text one piece at a time, predicting what comes next from what you gave them. When a task has several stages, that left-to-right habit can backfire: the model jumps to a final answer, skips the reasoning, or quietly drops a step you cared about. RISE works because it forces the order to the surface. By writing the steps down, you keep the model on a track you chose instead of letting it improvise. (For the bigger picture on why this matters, see why better prompts matter.)
RISE is a natural fit for students and educators, because so much of school is procedural. Solving a problem, running a lab, drafting an essay, grading against a rubric — these are all processes with a right order. A student who asks the AI to work a problem "one step at a time, pausing for me to try each part" gets a tutor instead of an answer key; a teacher who lays out the stages of an assignment gets a coherent activity instead of a pile of disconnected tasks. The rest of this guide breaks down each element with examples drawn from real studying, teaching, and research situations.
The four elements of RISE
Each card shows the question the element answers, with examples for studying, teaching, and research.
Role (R)
Who should the AI act as while it works?
- •Act as a physics tutor who walks me through problems step by step.
- •Act as an instructional designer building a lab activity.
- •Act as a methodical research assistant who shows its work.
Input (I)
What material should the AI work from?
- •Here is the full problem statement, including the given values and units.
- •Here are the three learning objectives the assignment must cover.
- •Here is my dataset and the question I am trying to answer.
Steps (S)
In what order should the AI proceed?
- •List the knowns, choose a principle, set up the equation, then solve.
- •Write the objective, then the materials, then the numbered procedure.
- •Pause after each step so I can check it before moving on.
Expectation (E)
What does the finished result look like?
- •Show every step with units, and state the final answer separately.
- •A one-page handout a student can follow without my help.
- •Flag any assumption you had to make to finish the task.
Steps vs. Expectation?
These two are easy to blur. The Steps are the path — the order of operations the AI should move through. The Expectation is the destination — what the final deliverable looks like and how you will judge it. Steps control the process; Expectation controls the product. Strong RISE prompts pin down both.
Why RISE produces better answers
Each element of RISE heads off a specific way that multi-step prompts go wrong. Role sets the depth and voice of the work: "act as a tutor who explains as you go" produces patient, visible reasoning, while "act as a careful lab designer" produces precision and an eye for safety. Input is the raw material — and it is the most common thing people forget. The model cannot grade a rubric it has not seen or solve a problem whose numbers you never gave it, so leaving the Input out forces it to invent the very details that should be yours.
Steps are the element that makes RISE what it is. Spelling out the sequence does two things: it stops the model from racing to the end and skipping the reasoning, and it makes the response checkable, because you can see whether each stage was actually done. Expectation then closes the loop by defining "done" — the format, the constraints, and the success criteria — so the model is not guessing whether you wanted a worked solution, a summary, or a finished handout.
You do not have to label the parts with the literal words "Role" and "Steps" — the model does not require them — but writing them out keeps you honest and makes sure the order is actually there. That is the real value of the framework: it pushes you to think through the process before you ask, which is often half the work. If you would rather not assemble the pieces by hand, the RISE prompt generator gives you a field for each element and writes the finished prompt for you.
RISE in action: before and after
The same goal, written two ways. Notice how much the model has to guess in the first version.
How do I solve this projectile motion problem? A ball is launched at 20 m/s at 30 degrees above the horizontal from ground level.
[Input] A projectile is launched from ground level with an initial speed of 20 m/s at 30 degrees above the horizontal. Neglect air resistance and use g = 9.8 m/s². I want to find the time of flight, the maximum height, and the horizontal range.
[Steps]
1. Resolve the initial velocity into its horizontal and vertical components.
2. Use the vertical motion to find the time to reach the highest point, then the total time of flight.
3. Use the vertical component to find the maximum height.
4. Use the horizontal component and the time of flight to find the range.
[Expectation] Pause after each step and ask me to attempt it first. Show every equation with units, keep the horizontal and vertical motions clearly separate, and state the three final answers together at the end.
The vague version usually dumps the whole solution at once, so you copy the answer and learn nothing about the method. The RISE version turns the AI into a tutor: by separating the horizontal and vertical motion and pausing at each step, it drills the one habit that makes or breaks projectile problems — treating the two directions independently — and you do the thinking instead of just reading it.
Make me a lab about pendulums.
[Input] The lab is for first-year physics students who have just learned about period and frequency. The goal is for them to discover how the period of a simple pendulum depends on its length. Available equipment: string, a few masses, a meter stick, and phone stopwatches.
[Steps]
1. Write a one-sentence objective and a testable question students will investigate.
2. List the materials and one safety note.
3. Write a numbered procedure that has them vary the length while keeping mass and release angle fixed.
4. Create a blank data table and 3 analysis questions that lead toward the length–period relationship.
[Expectation] Output a one-page student handout I can print. Keep the reading level appropriate for 9th grade, and make clear that only the length should change between trials.
Because the prompt fixes the role, the equipment, the learning goal, and the exact order of the build, the AI returns a coherent activity instead of a generic list — one with a controlled variable that genuinely isolates length, a printable structure, and analysis questions that point students toward the relationship rather than simply telling them the answer.
Common mistakes to avoid
- Forgetting the Input. RISE structures how the AI works, but it still needs the material. Paste the actual problem, the rubric, or the data — a perfect set of steps applied to missing information just produces a confident invention.
- Listing steps that are really just the goal restated. "1. Solve the problem. 2. Give the answer." is not a procedure. Good steps name the distinct stages — resolve, set up, solve, check — so the model cannot skip the part you care about.
- Putting the steps out of order. The model tends to follow your sequence literally. If "check the answer" comes before "solve," you will get exactly that. Read your steps in order and make sure each one could actually happen before the next.
- Confusing Steps with Expectation. Steps are the path; Expectation is the finished product. If you describe the output format inside your steps and leave the Expectation blank, the model often stops at the last step without ever shaping the final deliverable.
- Not asking it to pause. For studying, the whole point is to work it yourself. If you do not say "wait after each step for my answer," the AI helpfully solves everything at once — and the learning evaporates.
When to use RISE (and when not to)
RISE is the right tool when the task is a process — solving problems step by step, building a deliverable in stages, or running a procedure you want followed in order. When the process is not the point, one of the sister frameworks may fit better:
- →CRAFT (Context, Role, Action, Format, Task) — when you mainly care about the content and structure of a single, substantial output like a study guide or essay.
- →TAG (Task, Action, Goal) — for quick, everyday, one-off requests where laying out steps would be overkill.
- →CO-STAR — when the audience and tone of the message matter most, such as emails or announcements.
- →CARE — when you need the output in a strict, repeatable format every time.
Not sure which to pick? How it works walks through the whole methodology, and the use cases page shows the frameworks applied to real situations.
Frequently asked questions
What does RISE stand for in prompt engineering?
RISE stands for Role, Input, Steps, and Expectation. It is a structured prompt framework that tells the AI who to be, gives it the material to work from, lays out the exact sequence of steps to follow, and defines what the finished result should look like — which makes it ideal for any task you want completed as a defined, repeatable process rather than a single answer.
When should I use RISE instead of CRAFT?
Reach for RISE when the process matters as much as the result — working through a problem one step at a time, building a deliverable in stages, or running a procedure you want followed in order. CRAFT is better when you mainly care about the shape and content of a single output, like a study guide or an essay. The giveaway is the word 'Steps': if you can write a numbered list of what the AI should do, RISE is the right fit.
Is the RISE framework good for students?
Yes, especially for problem-solving. By giving the AI explicit steps — 'first identify the knowns, then choose the principle, then set up the equation, then solve' — you turn it into a guided tutor that walks you through the method instead of just handing you the answer. That is far more useful for learning, because you see the reasoning at each stage and can catch where your own thinking goes wrong.
Do I have to spell out every step myself?
Not always. If you already know the procedure, listing the steps gives you the most control. If you do not, you can ask the AI to propose the steps first and confirm them before it starts — for example, 'List the steps you will take to solve this, then wait for my approval.' Either way, making the steps explicit is what separates RISE from a vague request.
Which AI models does the RISE framework work with?
All of them. RISE is model-agnostic — the same structured prompt improves results in ChatGPT, Claude, Gemini, Microsoft Copilot, and open-source models. Every large language model produces more reliable multi-step work when it is given a clear role, the right input, an ordered procedure, and a definition of what done looks like.
Ready to try it out?
Open the builder, fill in the four fields, and copy your finished prompt into ChatGPT, Claude, or Gemini.