Singapore – Criteo has introduced an agent-based recommendation service intended to support the development of AI-enabled shopping assistants with more accurate and context-aware product suggestions derived from commerce data.
As large language models and retailer-built chat interfaces increasingly shape how consumers search for, assess and purchase goods online, the underlying recommendation systems are required to move beyond static product information.
AI shopping tools depend on access to behavioural and transactional signals in order to reflect real purchasing patterns rather than relying solely on publicly available descriptions. The newly introduced service is aligned with Criteo’s previously outlined approach to agent-based commerce.
“The real competitive advantage in agentic commerce will come from access to high-quality commerce data at scale,” Michael Komasinski, chief executive officer of Criteo, commented.
“This service brings that intelligence into AI-driven shopping experiences in a way that works for the entire ecosystem, delivering relevancy for consumers while respecting retailer data, brand integrity, and platform trust.”
The service is delivered through Criteo’s MCP and is designed to link AI shopping assistants directly with retailer inventories. It converts consumer queries into ranked product selections that are ready for transaction, using real-world shopping and purchasing signals that are not typically accessible through standard web indexing methods. This enables AI tools to present more tailored product options that take into account individual intent, popularity trends and stock conditions.
The recommendation system supports both open-ended browsing requests and specific product searches, while also identifying related items where relevant. Rather than returning full catalogues, it provides a narrowed set of options intended to simplify comparison and decision-making within AI-led shopping environments.
Criteo has been conducting trials of the service with an LLM provider since 2025 and is extending testing to additional AI platforms, retailers and brands as part of its broader evaluation of agent-based commerce applications.
