30 Gemini Prompts for Nursing Students (2026 Guide)

Nursing study has an unusual shape. The volume is enormous, the material spans physiology, pharmacology, law and communication, and most of it has to be recallable under time pressure rather than merely understood. That combination suits a tool that will quiz you indefinitely without getting bored.

It also contains one task where being approximately right is worse than being no help at all, and the research on that task is unambiguous. So this guide gives you 30 prompts organised by revision job, and it leads with the evidence, including the part that should stop you using the tool for one particular thing.

The boundary first. This is about studying nursing, not practising it. Nothing below is for use on a real patient, a real drug chart, or anything happening on a ward.

What the research says about Gemini and nursing

Start with the encouraging finding, because it is real.

A study published in BMC Nursing in May 2026 tested four models against 55 intensive care nursing questions drawn from midterm and final exams at a Turkish university. Gemini scored 78.18 percent, ahead of ChatGPT 5.0 Plus on 74.55 percent, ChatGPT 5.0 on 72.73 percent and DeepSeek on 61.82 percent. On the 22 questions rated difficult the spread widened, with Gemini on 77.27 percent against DeepSeek's 45.45 percent.

The caveat matters as much as the headline, and the authors state it plainly. The overall difference between models was not statistically significant, at p=0.296. So this is not evidence that Gemini is the best tool for nursing revision. It is evidence that the serious models sit in a similar band, and that band is somewhere around three correct answers in four.

Hold that number in mind, because it defines what the tool is for. One wrong answer in four is perfectly workable when you are generating practice questions and checking them against your course materials. It is close to useless when you are using it to find out whether your own answer was right, because the wrong answers arrive with the same confident explanation as the correct ones.

The one thing to never use it for

Dose calculation. Not as a shortcut, not as a check, not as a second opinion.

Three models were tested on computing maximum safe doses of local anaesthetics. ChatGPT recommended unsafe doses in 90 percent of cases, exceeding safe limits by an average of 198 percent. Copilot was unsafe in 67 percent, exceeding by 217 percent. Gemini produced the fewest unsafe recommendations of the three, which sounds reassuring until you read why. It frequently declined to give a figure at all and referred the question to a specialist. Where it did supply a dose range for mixtures, it exceeded safe limits by 140 percent.

The authors' conclusion is worth quoting rather than paraphrasing. These models "currently lack the accuracy and reliability needed for safe LA dose calculation" and "should not be used as decision-making tools for this purpose".

A separate study of drug dosage questions drawn from Harrison's Principles of Internal Medicine found Gemini's predecessor answering 54.76 percent correctly, against ChatGPT 4 on 83.77 percent. Barely better than chance, on the family of questions where a nursing student has the least room to be wrong.

Both studies used earlier model versions, so the numbers will have moved. The direction will not have moved enough to matter. Dose calculation has a right answer, a checkable method and a professional procedure attached to it, and none of those are improved by asking a language model to guess faster.

The boxes are blank on purpose. Working out what goes in them is the revision, and that is the job every prompt below is shaped around.

How to get a usable answer

A nursing prompt needs three things. It needs your year and programme, because the expected depth differs enormously between first year and finals. It needs the module or system, so the answer lands in the right frame. And it needs an instruction about what form you want, since the default is a tidy summary and a tidy summary is rarely what revision requires.

The prompts below assume that shape. Fill in the brackets and treat the first answer as something to check rather than something to memorise.

Pathophysiology and the why behind the condition

The strongest use, because explanation is what the tool does well.

1. Trace a condition from cause to presentation

I am a [year] nursing student. Walk me through [condition] from the initial cause to the signs and symptoms I would actually observe. At each step, tell me which observation links to which underlying change.

2. Connect the sign to the mechanism

Explain why a patient with [condition] presents with [sign or symptom]. I want the physiological mechanism, not the association, because I need to justify it in an exam answer.

3. Separate two conditions that look alike

Compare [condition A] and [condition B] for a [year] nursing student. Cover underlying mechanism, typical presentation, and the single observation most likely to distinguish them at the bedside. Put it in a table.

4. Work backwards from an observation

A patient's [observation, for example respiratory rate] is [value]. List the possible underlying causes a nursing student should consider, grouped by body system, and say which additional observation would help narrow it down.

5. Build the deterioration picture

Describe how a patient with [condition] typically deteriorates, in the order the changes usually appear. For each stage, tell me what I would see in the observations and what the expected nursing response is at [year] level.

Pharmacology, for learning the classes

Keep this to understanding drug groups and their effects. Calculations are covered above, and the answer there is no.

6. Learn a drug class properly

Summarise [drug class] for a [year] nursing student: how it works, what it is given for, the side effects I will be examined on, and the specific nursing observations required when a patient is taking it.

7. Build the monitoring list

A patient is prescribed [drug class]. What should a nurse be monitoring, how often, and what finding would mean escalating? Explain why each observation matters rather than just listing it.

8. Separate drugs that blur together

I confuse [drug A] and [drug B]. Build a table separating them by mechanism, main adverse effect, key nursing consideration, and the clinical situation where one is chosen over the other.

9. Work from the side effect backwards

Which commonly prescribed drugs cause [adverse effect]? Group them by the mechanism that produces it, and for each tell me how it would present in a patient I was observing.

10. Drill interactions and contraindications

Quiz me on the contraindications and significant interactions for [drug class] that a [year] nursing student is expected to know. One question at a time, and correct me before moving on.

Care planning and nursing process

Where the tool is useful for structure and weak on judgement, so use it for the former.

11. Structure a care plan

For a hypothetical patient with [condition], outline a care plan using [framework my course uses]. Give assessment priorities, nursing diagnoses, goals, interventions and evaluation criteria. Keep it at [year] level and label anything that needs individualising.

12. Prioritise when everything is urgent

Here are four hypothetical patients with these presentations: [list]. Rank who I should see first and explain the reasoning at each step. Then tell me what additional information would change the order.

13. Write measurable goals

Turn these vague nursing goals into specific, measurable, time-bound ones: [paste]. For each, say what evidence would show the goal had been met.

14. Rehearse the difficult conversation

Act as a patient who is [anxious, angry, or refusing care] in this hypothetical scenario: [describe]. Respond realistically to what I say. Afterwards, tell me what I handled well and what a mentor would have pulled me up on.

Clinical skills and documentation

15. Break a skill into steps

Break down [clinical skill] into the steps I would be assessed on, including the preparation and the checks before and after. Flag the three steps students most often miss.

16. Practise structured handover

Here is a hypothetical patient scenario: [describe, with no real details]. Write an SBAR handover for it, then critique my own version: [paste mine]. Tell me what I left out that the receiving nurse would need.

17. Tighten your documentation

Here is a practice entry I wrote for a fictional patient: [paste]. Assess it for clarity, objectivity and completeness at [year] level. Point out anything subjective, vague, or that could be read two ways.

18. Learn the rationale behind the policy

Explain why [procedure or precaution] is done the way it is. I want the underlying reason, the consequence of skipping each step, and the question a mentor might ask to check I understand it.

Exam and NCLEX-style practice

Use it to generate and drill. Verify the answers elsewhere, for the reason set out at the top.

19. Generate a practice set

Write [number] NCLEX-style questions on [topic] at [year] level, with four options each and plausible distractors. Present them one at a time, do not reveal answers until the end, then group my errors by topic.

20. Make it explain the distractors

For this question and its options: [paste]. Explain why each wrong option is wrong, and what misunderstanding would lead a student to pick it.

21. Drill prioritisation questions

Give me ten prioritisation and delegation questions at [year] level. After each answer I give, explain the principle that decides it rather than just confirming right or wrong.

22. Mark your written answer

Here is an exam question and my answer: [paste]. Mark it as an examiner would at [year] level, tell me the marks I missed and why, then write a full-mark version.

23. Build a revision timetable

My exam is in [time] and my weakest areas are [list]. Build a plan that weights the weakest high-yield topics, uses spaced repetition so earlier topics return, and leaves the final days for practice questions only.

24. Rehearse the viva or OSCE station

Act as an examiner at an OSCE station on [topic] for a [year] nursing student. Ask me questions one at a time, including follow-ups when an answer is thin. Give no feedback until the end, then tell me where I was vague.

Lectures, notes and papers

The safest category of all, because here the tool is transforming your own material rather than supplying facts.

25. Turn a lecture into recall questions

Here are my notes from a lecture on [topic]: [paste]. Turn them into 20 active recall questions and a one-page summary. Flag anything in my notes that looks incomplete or contradictory.

26. Make a paper readable

Summarise this paper for a [year] nursing student: [paste or attach]. Give me the question it asked, the method in two sentences, the main finding with the numbers, the limitation that matters most, and what it does not show.

27. Build flashcards from your own notes

Turn these notes into flashcards: [paste]. One fact per card, question on the front, answer on the back, formatted so I can paste them into Anki. Avoid cards that can be answered by recognising the wording.

28. Find the gap in your notes

Here are my notes on [topic]: [paste]. Compare them against what a [year] nursing student is expected to know, and tell me what is missing, what sits at the wrong depth, and what I can safely cut.

29. Teach it back and be corrected

I am going to explain [topic] as if teaching a first-year student. Tell me where I was wrong, where I was vague, and which part a patient would not have understood. Here is my explanation: [paste].

30. Plan a week around placement

Here are my placement days, shifts and deadlines: [paste]. Build a realistic study week that protects retrieval practice over rereading, puts the hardest topics on my best days, and assumes I will be exhausted after a long shift.

Placement, patient data, and the line

Two rules carry more weight than every prompt above, and both concern what you put in rather than what comes out.

Nothing that identifies a patient goes into a general chatbot. Not placement notes, not handover sheets, not ward lists, not drug charts, not clinical photographs, not anything with a hospital number on it. If you want to ask about a situation you saw on shift, rewrite it as a hypothetical with the details changed first. This is a confidentiality obligation and it does not soften at the end of a long shift.

And keep the clinical judgement where it belongs. The dose findings above are the sharpest illustration, but the principle is broader. These tools are weakest wherever an answer has to integrate several facts and carry real consequences, which is precisely the territory your mentors, your formulary and your course procedures exist to cover.

For the general-purpose version of this guide, see our general Gemini prompts for students, and if you are on a medicine programme the Gemini prompts for medical students guide covers anatomy and clinical reasoning with its own research findings. The prompt engineering basics explainer covers why these prompts are shaped the way they are, and why AI invents things is worth reading before you trust any of it with a fact you will be assessed on.

What this post does not cover

• Any clinical decision, dose, or action involving a real patient, which is outside the scope of a study guide and outside what the research supports
• Whether your university permits AI in assessed work, which is a question for your own programme's academic integrity policy
• Registration requirements, revalidation, or anything regulatory, since these vary by country and regulator
• Postgraduate or specialist practice, since the prompts here assume pre-registration study

Sources

• Comparative performance of artificial intelligence models in intensive care nursing questions, BMC Nursing, 2 May 2026 (55 questions; Gemini 78.18 percent, and the p=0.296 non-significance finding)
• Performance of 3 conversational generative AI models for computing maximum safe doses of local anaesthetics, JMIR AI (Gemini exceeding safe limits by 140 percent, ChatGPT unsafe in 90 percent of cases, and the authors' conclusion against using these as decision-making tools)
• Comparative evaluation of artificial intelligence systems' accuracy in providing medical drug dosages, PMC (Google Bard 54.76 percent correct against ChatGPT 4 on 83.77 percent)