easyai › ChatGPT for nurses

ChatGPT for nurses

Start with the rule, not the tool

Everything else on this page depends on one line, so it goes first.

That sounds like it rules out most of the job. It doesn't, and this is the part people get wrong in both directions. You can describe a situation in general terms without describing a person. “Write a handover note for a post-op patient who is stable but anxious about going home alone” contains no identifiers. It gets you the same wording help, with nothing to regret.

Some hospitals have an approved, contracted version of one of these tools. If yours has, the rules are different and your informatics team will tell you what they are. Assume you don't until someone in writing says you do.

What it's genuinely good at

Nurses don't need help nursing. The time sink is everything wrapped around it: the writing, the explaining, the same email for the fourth time. That's the part this helps with.

Making your charting read better, faster

Paste your own rough note, de-identified, and ask for it tightened into objective language. You are not asking it what happened. You already know. You're asking it to say what you wrote more cleanly.

Here are my rough notes on a shift. Rewrite them as a clear,
objective nursing note. Keep every clinical detail exactly as I
wrote it. Do not add anything I did not say. Flag anything that
sounds subjective so I can reword it myself.

[your de-identified notes]

“Do not add anything I did not say” is the important line. Without it, these tools smooth over gaps by inventing plausible detail, and a plausible invented detail in a clinical record is a serious problem. Read every line back before it goes anywhere near a chart.

Explaining things to patients and families

This is the strongest use, and the one nurses tell us saves the most time. Ask for a discharge explanation at a sixth-grade reading level, in the family's language, and you get a draft in twenty seconds that would have taken you fifteen minutes to write well.

Explain what a urinary catheter is and how to care for one at
home. Write it for an adult with no medical background, at about
a sixth-grade reading level, under 250 words, calm and practical.
No scare language. Then give me the same thing in Spanish.

You still check it. But checking a draft takes a minute and writing one takes fifteen.

The email you've rewritten four times

Staffing concerns, a scheduling swap, escalating something to a manager, a complaint you need to raise without sounding aggressive. Tell it who it's for and what you need to happen, and ask for it firm but professional. Getting the tone right on a difficult email is exactly the kind of thing it's good at, and exactly the thing that eats an evening.

Studying, and the NCLEX

Ask it to quiz you rather than explain to you. “Ask me one question at a time about fluid and electrolyte balance, wait for my answer, then tell me what I got wrong and why.” Being asked is how things stick. Being told isn't.

Two cautions. It gets exam-style questions roughly right but not always, so treat it as practice rather than as a question bank. And do not learn drug doses from it, for the reason below.

Where it will hurt you

• Drug doses, interactions and calculations. It produces numbers that look right and sometimes aren't, with no warning and total confidence. Every dose goes to your drug reference, your pharmacist, your protocol. There is no version of this where the AI is the source.
• Anything clinical about a real patient. It does not know your patient, cannot see them, and will answer anyway. It is a writing tool, not a second opinion.
• Policy and protocol. It will confidently describe “standard practice” that is not your unit's practice. Your policy is the policy.
• Citations. Ask for a reference and it may invent a real-sounding journal article that does not exist. If you need evidence, search for the evidence.
• Anything that becomes a legal record. Incident reports, statements, anything that could end up in a deposition. Draft those yourself.

The pattern behind all five: it's excellent at how to say something and unreliable at whether something is true. Keep it on the first job.

Try it now, on nothing that matters

Ten minutes, no patient involved, nothing to sign up for. Paste this into ChatGPT. It turns it into a tutor that gives you one step at a time and waits for you.

You are my patient tutor. I am a nurse. I am NOT technical.
Teach me by having me do one thing at a time, not by explaining.

Rules for this whole conversation:
- ONE step. Then STOP and ask what happened. Wait for my answer.
- No jargon about computers. Everyday comparisons only.
- If I say "it didn't work", ask what I saw. Do not move on.
- Keep every message under 150 words.
- Re-read these rules before every message you send me.

Lesson: write a patient handout I could actually use.

1. Ask me for one thing I explain to patients or families over and
   over. Just the topic, nothing about any real person.
2. Write it at a sixth-grade reading level, under 250 words, calm
   and practical. Show me.
3. STOP and ask me: "What's wrong with this, as a nurse?" There
   will be something. Get me to say it.
4. Fix that one thing without starting over. Show me the difference.
5. Remind me to check it against my own unit's guidance before I
   ever hand it to anyone.

Start with step 1. Ask me now, and nothing else.

Step 3 is the one that matters. You know more about this than it does, and the habit of correcting it instead of accepting it is the whole skill.

If it runs ahead of you, paste: Go back to being the tutor. One step at a time. I was on step 3.

Where to go next

• How to write ChatGPT prompts — the four things that turn a bland answer into a usable one.
• Use ChatGPT to make a plan — useful for nights and stretches of shifts you're dreading.
• The free ChatGPT course — six lessons, ten minutes each.

Questions