AI evidence extraction

Modelling the impact of generative artificial intelligence on information behaviour research

Thomas D. Wilson · Journal of Documentation · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10

Relevance

1/4

Quality (LMQS)

Evidence

Citations

0.00

FWCI

This is an AI analysis. Read the peer-reviewed original at the publisher: https://doi.org/10.1108/jd-03-2026-0150

Methodology & findings

Study design

Conceptual analysis grounded in a scoping review.

Sample

N = 89, 1 group

Primary method

Conceptual analysis grounded in scoping review methodology. No quantitative statistical analysis reported.

Main result

The study identified six critical characteristics of generative AI that require theoretical accommodation: "opacity, epistemic authority, sycophancy risk, personalisation, first-person voice construction and commercial optimisation." A modified model is proposed that introduces "ambient activation as a new activating mechanism; AI literacy, trust calibration and vulnerability status as new intervening variables; the AI system as an active intermediary node; ethical evaluation as a new stage between information processing and use; and bidirectional feedback loops representing individual re-prompting and macro-level knowledge distribution effects."

Reports effect sizes.

Research paradigm

Interpretivist/Conceptual

Author conclusions

The authors conclude that "the paper moves beyond applying existing information behaviour models to new contexts, proposing structural revision of Wilson's model to accommodate what generative AI systems are, rather than merely what users do with them."

Risk of bias

Scoping review methodology without explicit quality assessment criteria reported; Potential selection bias in paper identification and inclusion; No inter-rater reliability measures mentioned for study selection; Conceptual analysis lacks quantitative validation

Open questions raised

The paper identifies a fundamental theoretical gap: existing information behaviour models do not accommodate generative AI systems as active intermediaries. The authors propose structural revision to address this gap, suggesting future research should incorporate the identified characteristics (opacity, epistemic authority, sycophancy risk, personalisation, first-person voice construction, and commercial optimisation) into information behaviour frameworks.
The paper identifies theoretical gaps in existing frameworks that do not address generative AI as an active intermediary system. It notes that existing models of information behaviour assume interaction with passive retrieval systems and do not accommodate the characteristics of generative AI systems.
The paper identifies that "Existing models of information behaviour assume interaction with passive retrieval systems such as search engines and bibliographic databases. Generative AI systems, which actively generate personalised responses rather than retrieving existing documents, represent a fundamental shift that these models do not address."

Data: not_statedCode: not_statedExtracted from: pdfAgreement 80%

Explore related topics

Responsible AI Use Human-AI Collaboration Conceptual