From Search to Delegation: PhD Scholars’ Reliance on Agentic AI for Scholarly Information Discovery in Zambia

Context: Academic information discovery is changing as agentic AI systems identify, filter, summarize, compare, and recommend scholarly information. This rapid transformation raises questions about increasing reliance on AI and the continuing need for human oversight of these tasks.

Aim: The study aimed to explore how PhD scholars in Zambia use, delegate, rely on, and verify agentic AI for academic information discovery.

Approach: A quantitative, descriptive, cross-sectional survey was conducted at four Zambian universities. A total of 220 pen-and-paper questionnaires were distributed between 1 June and 25 July 2026, and 180 complete responses were received. The data were analyzed using frequencies, percentages, means, and standard deviations.

Key Findings: Participants used agentic AI for numerous information-discovery tasks, with initial summarization being the most commonly delegated task. Reliance was strongest for making scholarly discovery faster and easier. Participants perceived AI as useful and relevant, but confidence in its accuracy and credibility was lower. Verification was consistently high, especially for research claims, citations, bibliographic details, and source existence.

Interpretation: Delegation did not indicate unconditional dependence; rather, scholars combined AI-enabled efficiency with continued human verification.

Contribution: The study identifies a pattern of verified reliance and provides evidence of AI-mediated scholarly information behavior in the underrepresented Zambian and Global South contexts.