Nurse educators have always been magpies about content, and I mean that as a compliment. Long before anyone typed a prompt into anything, you were building orientation out of whatever was lying around: a unit policy, a colleague’s skills checklist, a journal article, a vendor’s product sheet, and a couple of decades of your own time at the bedside. You took what existed, worked out what was any good, and turned it into something your caregivers could actually use. AI is just the newest thing on that pile.

What’s changed is which part of the job is hard. For years the painful bit was making the material at all, the blank-page dread, the hours it took to get a first draft you weren’t embarrassed by. Now a free tool will turn a policy into a module in a few minutes. So the hard part has moved downstream, to checking the thing. And that checking, working out whether the content is accurate, current, and safe to put in front of a nurse, is the actual nursing work. It’s the reason AI is worth bothering with, and also the reason you can’t use it on autopilot.

This is written for clinical educators who are early in this, or still deciding whether to start. I’ll go through what AI is genuinely good at, the governance questions that matter more than tool choice, the intellectual property stuff people worry about (usually the wrong half of it), and how you get content out of a chat window and into the systems your learners actually touch.

AI is already in the room, mostly next door

In corporate learning and development, AI adoption has moved quickly. A 2025 survey of L&D professionals put usage at 87%. Another 2025 study, this one of 144 instructional designers, found 83% already using ChatGPT and about two-thirds reporting real time savings. Authoring-tool vendors now sell AI-assisted course building as several times faster than the old way, and they’re not entirely wrong.

Here’s the caveat, and it’s a big one. Narrow the lens to nursing, and specifically to educators using AI to build their own content, and the evidence gets thin fast. A scoping review that searched the literature up to early 2025 turned up only nine studies on nurse academics’ own use of generative AI, and pointed out that nearly all the attention so far has gone to what students do with these tools rather than what educators do. So take the numbers above as directional, borrowed from the field next door. Nobody has a settled picture of good practice in a hospital education department yet. You’re early. That’s genuinely fine, and it has an upside: the process that gets set up now is one you get to shape rather than inherit.

What isn’t up for debate is that the profession expects you to build skill here. In September 2025 the National League for Nursing put out a vision statement on AI in nursing education that treats AI literacy as a core competency and calls for policies instead of outright bans. So the direction is set. The specifics are for you and your colleagues to sort out.

What it’s good at, and where it will hurt you

Used well, AI takes the grind out of a first draft. It’s actually pretty good at turning a policy or a procedure into a sensible lesson outline, drafting quiz questions and scenarios for you to tidy up, rewriting dense clinical language into something a new grad can follow, and reshaping one piece of content into several (a microlearning refresher, an orientation module, a quick-reference card from the same source). It’s also excellent at the accessibility work that used to eat hours, like captions, alt text, and transcripts. For a team that’s short on time, getting to a decent draft in minutes is real relief, not a gimmick.

The danger is specific to your world. In most jobs an AI mistake is embarrassing. In a competency module it can be unsafe. These tools will invent a citation that looks completely real. They’ll state a medication detail or a guideline with total confidence and be wrong. They’ll reference a standard that doesn’t exist, or quote a version of one that got superseded two updates ago. And because the output reads so cleanly, the errors don’t wave at you. They sit there looking professional.

The line you hear everywhere is that AI content is “good enough with light editing.” For a marketing email, sure. For clinical content, “good enough” is exactly the trap, because a well-formatted module that’s subtly wrong is more dangerous than an obviously rough one. The rough one makes you slow down. The polished one invites you to skim and sign.

The questions worth asking before “which tool”

Most people start with which tool to use. These matter more.

Who’s accountable when it’s wrong? If an AI-drafted module teaches a step incorrectly and that reaches the bedside, the tool isn’t accountable for anything. A person is. Work out who that person is before you publish, not in the meeting after the incident.

Can you still call it evidence-based if you can’t trace the claim? Evidence-based means you can point at the evidence. If a nurse asks where a statement came from and the best you’ve got is “the AI said so,” it isn’t evidence-based, however authoritative it sounds. Traceability is most of the job.

What happens to your team’s own skill? There’s a real thread in the literature about deskilling, and it’s worth taking seriously. If drafting gets handed off entirely, the clinical judgment that makes a good preceptor good gets less practice. AI should be doing your educators’ typing, not their thinking.

And what are you actually feeding it? Every prompt is an input. Paste in a patient detail and you’ve got a privacy problem. Paste in a licensed textbook or a vendor’s protected content and you’ve got a different one. What goes in matters at least as much as what comes out.

Governance, or the approval problem

This is where most of the real work lives, and the good news is you don’t need to invent a new bureaucracy for it. You need AI-drafted content to go through the review you already run on any clinical material.

A workable model has two layers. First, a content and instructional review: does it teach clearly, does it hit the learning objectives, is it pitched right for the audience. Second, and this is the one you can’t skip, a clinical sign-off from a subject-matter expert who confirms it’s accurate and current. Both of these map onto things you almost certainly already have. Most hospital education departments run an education committee, some form of shared governance, or a policy-review process. AI content slots straight into those. The reviewer’s actual task doesn’t change much. What changes is the starting assumption: nothing is right until it’s checked, because a machine wrote the first pass and machines are confidently wrong.

Keeping it current is its own discipline, and arguably a harder one now, because it’s so easy to produce a mountain of content in an afternoon. Tie your review cycle to the source of truth, so that when a policy changes, everything built from that policy gets flagged for review automatically. Version it. Date it. Record who approved it and when. And make sure you can tell which version a given learner actually completed, because that’s the question you’ll be asked when it matters.

A few habits worth building in from the start rather than bolting on later: mark internally anything that began as an AI draft, so the next reviewer knows to look harder. Never let AI-drafted clinical content publish without a named clinical sign-off, and make that a rule rather than a strong preference. Feed the tool your approved policies and standards instead of the open internet, because the narrower and more trusted the source, the safer the draft. And keep the trail: what was made, who reviewed it, when it was last touched.

That last one connects directly to the thing that keeps you up at night. Joint Commission surveys and Magnet appraisals both come down to whether you can show your competency content is current, evidence-based, and run through a real process. A clean, dated, attributable trail is the difference between answering that confidently and scrambling for it. If AI multiplies how much content you produce, it multiplies what you have to be able to account for. Build the audit trail into the workflow, not afterwards when you’re reconstructing it from memory.

The IP question people ask, and the one they should

Two different issues hide under “intellectual property,” and educators tend to fret about the less dangerous one.

The question people ask is ownership. Can you copyright content your team made with AI? Under current US guidance, purely AI-generated material isn’t copyrightable, and detailed prompts on their own don’t make you the author. Only the human contribution, the selection, the arrangement, the real editing your educators do, is protectable, and a federal appeals court held that human-authorship line in early 2025. So if you want to own and defend your organization’s competency content, the human authorship has to be real and documented. That’s US law, it’s still moving, and none of this substitutes for your own institution’s counsel.

The question that will actually bite you is on the input side. Don’t feed copyrighted or licensed material into a tool you don’t have the rights to use that way, and don’t assume a tool’s terms of service line up with your assumptions. Some tools reserve the right to train on what you upload. Others promise your content stays yours. Read the terms before you paste anything sensitive, and never paste patient data, full stop. For most education teams the licensed-content and privacy questions are a bigger day-to-day risk than the ownership one.

Getting it out of the chat window

A draft sitting in a chat window isn’t training. It has to get into the tools your learners use, and there are broadly two ways there.

The first is the one most teams already know. Draft in a general tool, then move the content into an authoring platform like Articulate Rise, where it becomes a module you can export to your LMS or competency management system. Rise has its own built-in AI now that generates outlines, blocks, quizzes, and images. One thing worth knowing: reviewers have noted that Rise’s AI doesn’t apply Bloom’s Taxonomy or any other learning model, so the quality of what it produces leans heavily on the clarity of the objectives you hand it. Vague objectives in, vague module out.

The second path is newer… There’s a category of AI-native tools now that will take an uploaded policy, a medical device manual, or a PowerPoint deck and turn it into a course for your LMS or competency system in minutes, with only light editing. These products are real, they’re getting better, and for simple content they save serious time. But apply one test before you commit to any of them: a slick auto-generated module still needs a clinical sign-off. If anything, mass-producing content multiplies the review load rather than removing it. A tool that spits out ten modules before lunch hasn’t saved you anything if it’s handed you ten modules that each need a preceptor’s eyes.

Whichever path you take, keep in mind what the content is actually for. A finished module proves someone completed training. It doesn’t prove they can do the thing. That gap, between “watched the module” and “demonstrated it at the bedside,” is the one that surfaces in a survey and, more importantly, on the unit. It’s why competency content is only ever half the job. Verifying the skill in the flow of work is the other half.

What good looks like

If you’re starting out, a short set of processors will carry you a surprisingly long way. Start from a clear learning objective rather than a blank prompt. Feed the tool your approved sources, not the open web. Require a named clinical sign-off on everything before it publishes. Disclose and date what was AI-assisted. Version content and review it on a cycle tied to your policy revisions. And protect your inputs: no patient data, no licensed material you don’t have the rights to.

This is the same judgment you’ve always applied to curriculum content, just pointed at a faster source.

Which is roughly where this started. Nurses were never good at building learning because they were quick writers. They were good at it because they knew what good looked like, what mattered on the unit, what a new grad needed, and what would hold up when someone came looking. AI can draft all day. Judging is the part it can’t do, and that’s exactly the part you bring. It’s also what makes AI safe to put to work in the first place.

Working through governance and audit-readiness for your competency content? That’s the problem Xapify was built to help with. Start a conversation with our team.