West African doctors were willing to use proven AI—but most questioned its reliability and autonomy costs
A survey of 136 physicians and 72 key-informant interviews found high conditional willingness alongside low experience, reliability concerns and demands for local validation. The sample was predominantly Nigerian and measured views, not patient outcomes.
The short version
- What happened
- Among 136 physicians affiliated with the West African College of Physicians, 94.1% said they would use AI if it were proven effective, but only 33.1% had used AI tools.
- Who it affects
- Clinicians want evidence that a tool works with local patients and infrastructure before it becomes part of routine care.
- What remains unproven
- The survey was cross-sectional and measured expectations, concerns and willingness rather than behaviour, diagnostic performance or patient outcomes.
Keep in mind
The survey was cross-sectional and measured expectations, concerns and willingness rather than behaviour, diagnostic performance or patient outcomes.
Participants were predominantly Nigerian and affiliated with one professional college; regional and profession-wide representativeness is limited.
Percentages describe the surveyed physicians and should not be generalized to all West African clinicians.
The study received no direct research funding. The authors declared no competing interests.
Source analysis · 1 linked source. Primary analysed source date: 31 October 2026.
Read the full report · 3 min The evidence, methods, implications and limitations
Willingness was conditional, not blanket acceptance
Physicians connected to the West African College of Physicians saw possible clinical value in AI, but did not equate promise with trust. In a cross-sectional survey of 136 doctors, 85.3% agreed that AI could improve diagnostic accuracy and 83.1% thought it could reduce errors. Only 33.1% had used AI tools, however, while 67.6% rated AI information as unreliable and 77.9% saw a threat to clinical autonomy.
The apparent tension matters more than a simple pro- or anti-AI label. Some 94.1% said they would use AI if it were proven effective. Trust was the strongest measured predictor of willingness (beta 0.48), and previous use also had a smaller association (beta 0.21). Because this was a one-time survey, those relationships do not show that raising trust would cause adoption.[1]
Local validation came before scale
Across 72 key-informant interviews, participants described AI as a possible cognitive partner for decision support, error reduction and administrative relief. They also raised over-reliance, skill erosion, workflow disruption and unclear accountability. Their recurring conditions were transparent reasoning, testing in similar patient populations, clinician-centred design and independent oversight.
Those conditions are especially relevant where infrastructure, specialist staffing and locally representative data are uneven. Yet the sample was predominantly Nigerian and drawn through a professional college, so it cannot stand in for every profession, health system or country in West Africa. The next evidence should put specific tools into real workflows and measure errors, time, access, clinician behaviour and patient outcomes—not only stated willingness.[1]
The distinction that matters
What clinicians said would turn interest into safe use
High willingness depended on evidence and governance that fit local clinical conditions.
Prove performance
Validate in similar patients and workflows
94.1% said they were willing if effectiveness was demonstrated
Preserve judgement
Keep clinician-centred oversight and explainability
Addresses autonomy, over-reliance and skill-erosion concerns
Define responsibility
Set independent governance and accountability
Clarifies who acts when an AI-supported decision fails