Your supervisor’s first reaction to “AI in nursing education” as a topic is usually the same one: too broad, too trendy, not clear what you would actually measure. That reaction is fair — the phrase names a field, not a research problem. Every extra week spent circling a vague version of this topic is a week closer to a registration deadline or a lapsed bursary, and a topic your committee sends back once already costs you a full proposal cycle to fix.
Why does “AI in nursing education” get rejected as a topic on its own?
Because it names a technology and a field, not a problem. A defensible thesis needs a specific population (which students or educators, at which level), a specific use of AI (a chatbot for exam revision, an AI-assisted clinical simulation tool, generative AI for care-plan writing practice), and a specific outcome you can actually measure (readiness to adopt it, perceived usefulness, an effect on a learning outcome, or educators’ attitudes). Narrowing those three things turns a vague trend topic into a research problem your supervisor can approve in one meeting instead of three rounds of revision.
How do I write a problem statement that isn’t just a hunch?
A workable structure: state the trend (AI tools are increasingly available to nursing students and educators), state what is documented so far in the literature (general AI-in-higher-education adoption research, and any nursing-specific studies you have actually read), and state precisely what remains unaddressed for your specific population, institution type or AI use case. If you cannot point to the specific gap in one sentence, the problem statement is not ready to write yet — go back and narrow the population or the AI use case further before drafting Chapter 1. This is exactly the same disciplined narrowing process a genuine research gap requires in any field, not a shortcut specific to a trend topic.
What does the wider literature already say about AI adoption in nursing education?
International nursing education literature on AI adoption has grown quickly in recent years, generally covering three areas: student and educator attitudes toward AI tools, the use of AI in clinical simulation and skills training, and integrity concerns around generative AI in academic writing and assessment. Rather than citing a specific figure or study you have not personally opened and verified, describe this literature at the level you can actually defend — “a growing but still limited body of international literature has begun to examine X” is honest and defensible; a specific adoption percentage attributed to a study you have not read is not. Read the actual papers closest to your chosen population and AI use case before drafting your literature review, and let their real findings — not an assumed consensus — shape your gap statement.
What theoretical framework fits a study like this?
Technology-adoption models are the standard anchor for a study on whether and how nursing students or educators take up an AI tool — the Technology Acceptance Model (Davis, 1989) and its extension, the Unified Theory of Acceptance and Use of Technology (Venkatesh et al., 2003), are the two most commonly cited in this kind of study internationally, measuring constructs like perceived usefulness, perceived ease of use, and behavioural intention to use a new technology. If your study is more about educators’ concerns or ethical reservations than adoption itself, a qualitative design without a formal adoption model — built instead around your own thematic categories — may fit the research question better. State which route you are taking, and why, in the theoretical framework section rather than naming a model you are not actually going to apply.
A worked illustrative example
This example is entirely illustrative and fictional. Working title: “Nursing Students’ Perceived Usefulness and Readiness to Use AI-Assisted Clinical Case Simulations: A Survey at a South African Nursing College.” Problem statement: AI-assisted clinical simulation tools are increasingly marketed to nursing education programmes, but no South African study has measured how ready and willing undergraduate nursing students actually are to use them, or what factors predict that readiness. Primary objective: to determine nursing students’ perceived usefulness of, and readiness to use, AI-assisted clinical case simulation tools. Secondary objectives: (1) to describe students’ current exposure to AI tools in their coursework; (2) to determine whether year of study and prior digital-literacy confidence predict readiness; (3) to identify students’ stated concerns about using AI-assisted simulation for clinical learning. Research questions mirror each objective directly. Theoretical framework: the Technology Acceptance Model, adapted to name the specific tool and population rather than “technology” generically. Method: a cross-sectional survey, analysed with descriptive statistics and a regression testing which factors predict readiness.
What method fits this kind of study?
A cross-sectional survey of students or educators is the most common design for an adoption-focused study like the example above, since it can reach a reasonable sample size within a typical honours or master’s timeline. A qualitative design — interviews or focus groups with nurse educators about their concerns, workload implications or perceived risks — fits better where the research question is about meaning and experience rather than predicting adoption. A mixed-methods design (survey plus a smaller set of follow-up interviews) is common where a study wants both the breadth of a readiness score and the depth of why educators feel the way they do. Choose based on your actual research question, not on which design sounds more rigorous on paper.
What common mistakes weaken this kind of proposal?
Three recur most often. First, keeping the topic at the technology level throughout the whole proposal rather than narrowing it in the title, objectives and questions consistently — a narrowed problem statement followed by a generic “AI in education” literature review undoes the narrowing work. Second, choosing a theoretical framework because it is well known rather than because it fits the actual research question — TAM fits an adoption or readiness question; it does not automatically fit a study about ethical concerns or curriculum policy, which needs a different framework or none at all. Third, treating “students’ attitudes toward AI” as inherently interesting without connecting it to a concrete decision the findings would inform — naming who would use the results (a specific department deciding whether to introduce a tool, for example) sharpens the significance section considerably.
How is this different from asking whether AI is allowed for writing the thesis itself?
This is a common point of confusion worth stating explicitly in your own proposal: a thesis about AI in nursing education studies AI as its subject matter — how students or educators perceive, adopt or resist a specific tool. That is a completely separate question from whether you are permitted to use an AI tool while writing your own dissertation, which the site’s guide to AI use in South African dissertations covers. A study that studies AI adoption in nursing education, using entirely conventional research methods to design, collect and analyse its own data, has nothing inherently to do with how the write-up itself was produced.
Where does this connect to the rest of your proposal?
Once your problem, objectives and theoretical framework are set, the population and sampling section should specify exactly which students or educators, at which institution or programme level, and the ethics section needs to address informed consent and, where students are surveyed by their own lecturer, the specific steps taken to avoid coercion — a standard concern in any study where a researcher has some authority over participants. The site’s guide to ethics approval and informed consent for a nursing research report covers this in more depth, and if your design uses a validated instrument for a related construct (digital literacy, self-efficacy), the site’s guide to choosing a validated scale for a nursing research report is the next step. If you are still deciding between this topic and another nursing angle entirely, the site’s mental health nursing research-gap examples piece shows the same narrowing process applied to a different sub-field.
What if my supervisor still says the topic is “too trendy”?
This objection almost always means the topic is still stated at the technology level rather than the research-problem level. Bring a one-paragraph version narrowed to a specific tool, population and measurable outcome — like the worked example above — rather than defending the general importance of AI in healthcare education. A supervisor who sees a specific, testable question with a named theoretical framework and method is evaluating a research proposal; a supervisor who sees “AI in nursing education” alone is evaluating a headline, and headlines get sent back.
Structuring a topic like this properly — narrowing the problem, aligning objectives with a real method, keeping the theoretical framework honest — is exactly where Tesify helps, turning a vague trend into a proposal your supervisor can actually approve, while every finding and every word stays entirely yours.
Frequently asked questions
Is “AI in nursing education” too saturated a topic to be accepted?
The general area is popular, but a specific, narrowly scoped version — a named tool, a named population, a measurable outcome — is not automatically saturated just because the broad topic is trending. Narrowing is what makes it defensible, not avoiding the area entirely.
Do I need access to an AI tool myself to study this topic?
Not necessarily — a study of perceptions, readiness or attitudes can be conducted through a survey or interviews about a tool participants already use or are aware of, without you needing to build, license or personally operate the AI system yourself.
Can this topic work for an honours-level research report, or only a master’s dissertation?
It can work at either level — an honours-level version would typically use a smaller, more descriptive design (for example, only the perceived-usefulness objective, without the predictive regression), while a master’s-level version can support the fuller worked example above.
Will using a well-known theory like TAM make my study seem unoriginal?
No — applying an established, validated framework to a new population or context (South African nursing students specifically, rather than a generic student sample) is a legitimate and common form of originality in applied education research; the originality is in the context and findings, not in inventing a new theory from scratch.
What if my institution doesn’t have any AI tools in use yet to study?
You can still study perceptions and anticipated readiness before adoption, or study educators’ policy and planning decisions around a tool they are considering — the absence of current use is itself a valid starting point for a “readiness” or “anticipated adoption” framing rather than a barrier.
How long would a study like the worked example realistically take?
A single-institution cross-sectional survey of the kind in the worked example is generally achievable within a standard honours or one-year master’s research timeline, provided ethics clearance and data collection do not face unusual delays — build in buffer time for both, since they are the two most common sources of delay in this kind of study.