This post was originally published on this site.
Criteo has set out a set of expectations for how agentic AI will shape digital commerce, with a focus on search fragmentation, retailer-owned shopping assistants, and advertising-led monetisation for large language model platforms.
Michael Komasinski, Chief Executive Officer, Criteo, said the shift will play out as an addition to existing routes to purchase rather than a replacement for established channels.
“Agentic commerce will follow the same trajectory, adding a powerful new layer of engagement, but not a wholesale replacement for existing ones,” said Michael Komasinski, Chief Executive Officer, Criteo.
Incremental channel
Komasinski positioned agentic commerce as another surface for consumer engagement. He compared the likely impact with earlier technology shifts in retail, including eCommerce, mobile browsing and social commerce.
He cited third-party forecasts that keep agentic activity as a minority share of total retail sales in the medium term. He pointed to global eCommerce at 20% of total retail sales. He also referenced a projection that 25% of eCommerce could be agentically driven by 2030. He said that would translate to slightly above 5% of total retail sales.
Komasinski said consumers still want control over purchase decisions. He said current tools perform best in research and comparison. He said uptake will track how well systems reduce friction in shopping journeys, including time spent searching and effort spent narrowing options.
Discovery battle
Criteo’s outlook also emphasised the impact of agentic assistants on product discovery. Komasinski said search has already fragmented across multiple environments. He cited community platforms such as Reddit as one example of an alternative to traditional search.
He said prompt-based queries on large language model platforms add another discovery layer. He said this change spreads attention across more touchpoints. He said product findability will become a competitive focus across retailer sites and apps, social platforms, and AI-driven environments.
Criteo referenced its own consumer research. It said a global survey of 10,000 respondents found 40% of US shoppers use agentic shopping assistants regularly for product research. It also said 96% of those shoppers use other channels along the way, including search engines, social platforms, and brand and retailer sites.
“Products must be findable wherever consumers initiate queries: across retailer front ends, social platforms, LLM platforms, and other emerging AI-driven environments,” said Komasinski.
Retailer front ends
Criteo said retailers will face pressure to modernise customer experiences across websites, apps and in-store journeys. It said consumers increasingly expect guided discovery and conversational interfaces. It also said these expectations reflect experiences people have become accustomed to on large language model platforms.
Komasinski challenged the idea that large language model platforms will take revenue away from retail media networks. He said retailer environments remain central to conversion, fulfilment and loyalty.
He cited examples of retailer-owned assistants. He pointed to Sensor Tower data on Amazon’s Rufus assistant. He said Rufus-assisted Amazon sessions resulting in a purchase grew more than 100% during Black Friday and Cyber Monday. He contrasted that with 20% growth for non-Rufus sessions. He also referenced Walmart’s Sparky assistant and said it is showing growing traction.
He also pointed to Accenture research. He said US shoppers prefer retailer or brand-specific chat assistants rather than third-party large language model platforms while shopping.
Komasinski said a commercial opportunity is emerging around sponsored recommendations inside retailer-owned chatbots. He also pointed to retailer app integrations on platforms such as ChatGPT. He said this model keeps retailer control over product ranking.
Ads for LLMs
Criteo argued that advertising will become the main monetisation route for large language model platforms. Komasinski said the question of sustainable revenue models will remain central as usage grows.
He cited Google showing ads in its AI Mode search engine as an early example. He contrasted advertising with affiliate and marketplace models. He said those approaches depend on capturing transactions and limit scale and interoperability. He said advertising monetises attention and intent across discovery and decision phases.
“The most sustainable and flexible model will be advertising, and we’re already seeing that play out today as Google has begun showing ads in its AI Mode search engine,” said Komasinski.
He also referenced ChatGPT’s subscription mix. He said only about 5% of its users pay for subscriptions, and he said ads will be the “logical engine” for growth for the free tier.
Data quality
Criteo’s final prediction focused on the limits of current AI shopping experiences without structured, current commerce data. Komasinski described a search for bicycle tyres in New York City and said the assistant returned broken links, discontinued products and incomplete information.
He said recommendations fail without high-quality, structured, real-time commerce data. He said gaps remain around inventory, pricing, product attributes, checkout and fulfilment systems. He also referenced an analysis from OpenAI and said ChatGPT shopping research delivers 64% accuracy.
Criteo said its data footprint gives it an advantage in this environment. It said it sees 720 million daily active users interacting across 4.5 billion SKUs in its Universal Product Catalog. It said this activity generates more than $1 trillion in eCommerce transactions each year. The company said it maintains detailed metadata across global retail feeds. It said this approach allows systems to surface products that are in stock and accurately described.
Komasinski also pointed to interoperability efforts across AI systems. He cited Criteo’s Model Context Protocol work and said it “enables interoperability across systems”.
“The future of agentic commerce will be led by those that deliver accurate, trustworthy, and interoperable pathways to product discovery,” said Komasinski.