What is it about?
This survey reviews how generative AI can help recommender systems make better use of context. A recommendation is rarely based only on what a person liked before; it can also depend on time, location, recent activity, goals, device, social setting, or even hidden factors such as mood and intent. We organize these contextual factors along three dimensions — whether they change over time, whether they are directly observable or hidden, and whether they are recorded as explicit features or learned by the model — and use this structure to compare how each family of generative models handles them. We organize eight families of generative models, including the same broad class of methods behind tools such as ChatGPT and image generators, and explain how each family can support context-aware recommendation.
Featured Image
Photo by Denise Jans on Unsplash
Why is it important?
A central question for researchers and practitioners is which generative method to choose when a recommendation problem involves fast-changing user behavior, sparse data, hidden contextual factors, or strict requirements on speed, interpretability, and deployment cost. This survey addresses that question by providing a unified taxonomy that links types of context to suitable model families. It also offers side-by-side comparisons, together with selection guidance. As generative AI becomes increasingly relevant to recommender systems, this work provides a timely reference for newcomers seeking orientation and experienced researchers identifying open problems.
Perspectives
One question I keep returning to is whether generative models will eventually replace the classical recommender-system pipeline, or whether they will remain one component within larger hybrid systems. The evidence we gathered suggests the latter for now: generative models are powerful, but they still need to be combined with retrieval, ranking, evaluation, feedback, and deployment infrastructure. At the same time, the boundaries are moving quickly, and I expect some parts of this survey to need updating sooner than a static article would ideally allow.
Alì Ghasempouri
Universita degli Studi di Bologna
Read the Original
This page is a summary of: Generative Models for Context-Aware Recommender Systems: A Survey, ACM Computing Surveys, May 2026, ACM (Association for Computing Machinery),
DOI: 10.1145/3816031.
You can read the full text:
Contributors
The following have contributed to this page
