Why the work stays at the bedside
Whether AI will replace nurses is really two questions: can software do nursing tasks, and can anything do the parts that happen in a room with a sick person. The first answer is partly yes. The second is mostly no. Registered nurses give medications and watch for reactions, assess a patient’s breathing, skin and pain, and decide when a small change means something serious. That work is physical, legally accountable and done under time pressure.
By our read of the task list above, 86% of task time still needs a person. The reason is not that nursing is mysterious. It is that most nursing tasks combine a judgment call with a pair of hands and a license behind them. A model can suggest that a dose looks wrong. A nurse has to hold the syringe, check the patient and answer for the result.
There is a second reason the job holds. Demand is large and rising. The BLS counted 3,379,720 registered nurses in the United States with median pay of $97,550, and projects employment to grow 5.6% from 2025 to 2035 (BLS, 2025). Tools that save charting time in a short-staffed unit tend to be absorbed as relief, not turned into vacancies.
What AI does, what it helps with, and what it leaves to nurses
The narrow slice AI can handle on its own is paperwork. Pulling vital signs and medical information into the chart, and keeping records and reports in order, are the clearest cases. Our coverage figure for this job is 14 out of 100, and the share of task time AI can do without a nurse in the loop is 0%. Ambient documentation tools sit here. How coverage is measured explains what counts as task time.
Assistance covers more ground: 14% of task time. Monitoring and reporting changes in a patient’s condition is one example, where alerting systems flag a trend early and the nurse decides what it means. Patient and family education is another. A draft handout or a translated explanation saves time, but the teaching still happens face to face.
Everything else stays with people. Giving medications and watching for side effects, preparing patients for and assisting with treatments and procedures, and calming someone who is frightened are not tasks AI takes over. They need touch, timing and someone who can be held responsible.
What has actually been tested
Not much, in this job. Our evidence grade here is D, which means there is no direct, published test of AI against registered nurses doing nursing work. So we publish no quality score for nursing. A grade is not a verdict; it is a statement about how much we know.
What would settle it is specific: trials in real units comparing AI-assisted documentation or triage with nurse-led practice, measuring charting time, missed changes in condition, medication errors and patient outcomes, reported with enough detail to check. Until that exists, claims that a system matches a nurse rest on demos, not measurement. Our rules for that question are on the quality parity page, and the wider method is set out in how we score jobs.
Good to know: the panel further down this page shows what leading AI assistants say about nursing, which is useful context but is not evidence of capability.
When this could change
Most likely after 2042 (8 in 10 of our scenarios). The replacement-year method explains how that window is built and what it does and does not claim.
Two things could pull it earlier. Documentation and monitoring software is cheap next to staffing, so it spreads quickly once a hospital group signs. And virtual nursing models move admissions, discharge teaching and chart review off the floor, which concentrates the hands-on work into fewer hours per patient.
Two things hold it back. The physical portion of nursing would need dexterous humanoid hardware able to move patients, place lines and manage dressings in cramped rooms, and that equipment is not working in wards at scale. Second, accountability sits with a licensed person. State practice rules, employer liability and consent all assume a nurse made the call. Changing software is fast. Changing who can legally be responsible is slow.
How to stay needed in nursing
Lean into the work that sits on the human side of the split. First, assessment: being the person who notices the change in a patient’s condition before the monitor does. Second, medication safety, including catching the order that does not fit the patient in front of you. Third, procedures and treatments at the bedside, from wound care to line management, where skill and judgment arrive together.
Two skills raise your value as the tools arrive. One is reviewing machine output critically: knowing how an ambient note or an alert can be wrong, and documenting the correction. The other is teaching and coordination, the handoff and family conversation that keeps a plan from falling apart between shifts. Nurses who can lead that work also tend to get a seat when a unit picks its tools.
If you are weighing paths, the closest roles are worth comparing side by side. See critical care nurses, nurse practitioners and licensed practical and licensed vocational nurses, or put any two jobs next to each other with our job comparison tool.
For wider context, the healthcare practitioners and technical family shows how nursing sits against other clinical roles, and the hospitals sector page covers the setting most nurses work in. Jobs built around in-person judgment cluster together, as the list of jobs that mostly need a person shows. Our score for nursing is 80 out of 100 (higher is safer).