02 Research evidence series
Our papers review research on AI in marketing, work and learning. Each one compares study findings, explains the methods and links to the original sources.
How AI assistance affects creative quality, collective diversity and subsequent unaided performance.
18 references · 3 figures · 20 min read
Evidence on generated advertising, audience response and why creative ratings or clicks do not establish incremental sales.
12 references · 3 figures · 16 min read
A review of AI disclosure, credibility and trust, separating consumer experiments from technical provenance and legal categories.
12 references · 3 figures · 16 min read
Compares synthetic survey validation across averages, rankings, subgroups and individuals, including recent polling evidence.
12 references · 3 figures · 15 min read
A synthesis separating measured task gains, workflow productivity, realised cost savings and wider economic returns from AI.
12 references · 3 figures · 18 min read
Research on project failure, task fit and adoption explains why technical demonstrations and sustained organisational use differ.
12 references · 3 figures · 16 min read
Task-based labour research and AI field experiments distinguish exposure, assistance, reallocation and observed employment outcomes.
12 references · 3 figures · 18 min read
Agent research distinguishes successful demonstrations from repeatable workflow completion under tool, policy and resource constraints.
12 references · 3 figures · 21 min read
A review of indirect prompt injection, model-side defences and the permission boundaries around tool-using agents.
13 references · 3 figures · 16 min read
Technical evidence on memorisation and extraction, alongside the processing distinctions needed to assess marketing data flows.
13 references · 3 figures · 16 min read
An evidence-led comparison of model openness, transparency and governance responsibilities across deployment arrangements.
13 references · 3 figures · 15 min read
Research on model scale, specialisation and compression, with a framework for testing quality-preserving substitution.
14 references · 3 figures · 18 min read
A technical review of inference memory, quantisation, serving performance and resource accounting for local and hosted systems.
14 references · 3 figures · 21 min read
A review of generative-search visibility research, citation quality and the limits of applying experimental gains across platforms.
13 references · 3 figures · 15 min read
Research on AI-supported strategic decisions, professional performance and the limits of inferring firm outcomes from rated plans.
12 references · 3 figures · 16 min read
Research on offloading, explanations, automation bias and interface design in human–AI decision-making.
12 references · 3 figures · 15 min read
A review of inference systems and unit economics, separating token prices from the full cost of accepted work.
12 references · 3 figures · 16 min read
A review of the engineering obligations that persist after an AI model is acquired or a working demonstration is built.
12 references · 3 figures · 15 min read
A review of why task productivity, agency cost, contract price and client value require separate evidence.
12 references · 3 figures · 16 min read
A detailed review of the measurement evidence behind organisational AI readiness and the limits of composite scores.
12 references · 3 figures · 21 min read
A review of heterogeneous human–AI outcomes, separating access and reliance from productivity and independent capability.
12 references · 3 figures · 17 min read
Tutoring experiments and learning-science evidence on scaffolding, feedback, retrieval and independent performance.
12 references · 3 figures · 18 min read
Evidence on shopping-agent selection, recommender concentration, consumer reliance and the limits of commercial inference.
12 references · 3 figures · 16 min read
Reviews hallucination measurement, calibrated uncertainty and verification, explaining why repeated agreement cannot establish truth.
12 references · 3 figures · 16 min read
Examines why prompt gains depend on task, examples and evaluation, with detailed treatment of robustness and instruction density.
13 references · 3 figures · 21 min read
Brand-memory research on identification, salience and distinctive assets, with implications for evaluating scaled content.
12 references · 3 figures · 20 min read
A review of benchmark coverage, behavioural testing, factuality and automated judges for evaluating AI-supported marketing work.
12 references · 3 figures · 16 min read
Explains retrieval grounding through architectures and benchmarks, separating relevant passages, faithful answers and citation support.
12 references · 3 figures · 18 min read
Studies of personalised advertising and generated messages, separating relevance, trust, response and incremental outcomes.
12 references · 3 figures · 16 min read
How experiments and modelled counterfactuals identify marketing effects, and why prediction and attribution do not establish incrementality.
12 references · 3 figures · 16 min read
30 papers
Dates show evidence cutoffs · Source references included in every paper
Try another title or topic, or clear the search.