Salesforce: Agentic search is becoming the first step in online shopping, with AI referrals surging

More shoppers are starting their buying journeys by asking an AI, rather than clicking through to a retailer or running a traditional web search. In Salesforce’s latest State of Commerce findings, the company says the use of agentic search as the first step in shopping grew 200% year over year, as discovery increasingly happens inside AI chats or retailer-run AI assistants before consumers reach a product page.

Salesforce ties the change to two pressures building at once. Eighty-six percent of commerce leaders say AI is raising customer expectations, and they are being asked to respond with fewer resources. In that context, expanding or implementing AI is described as the top priority for the year ahead, and it is also cited as the leading anticipated challenge.

The report’s behavioral evidence focuses on shoppers’ real activity across commerce sites. Salesforce says traffic referred from AI chats rose between 150% and 428% year over year in every quarter it measured, while overall traffic growth stayed in the single digits to low double digits. The company frames this as a sign that AI is not just influencing decisions, but also changing where shoppers begin.

Consumers’ self-reported behavior points in the same direction. Salesforce reports that between August 2025 and May 2026, discovery through brand-owned properties fell 7%, and traditional search declined 15%. Over the same period, the share of consumers choosing newer channels, including AI assistants, grew 38%.

Salesforce also says the next 12 months will hinge on whether commerce teams can prepare their product information for conversational discovery. It reports that 86% of respondents agree LLMs will be essential to product discovery within the next year. Common responses, according to Salesforce, include improving product content quality, optimizing content for conversational queries, and submitting data feeds to AI search platforms.

The company’s research suggests the shift is still concentrated in the discovery stage. Salesforce characterizes fully autonomous purchasing as early, even as the first conversation increasingly determines what consumers see and what they assume about a product by the time they arrive on a retailer’s site.

Agentic AI adoption is moving from pilots to broader rollout. Salesforce says fewer than a third of organizations currently use agentic AI, but nearly half of non-adopters plan to deploy it within the next six months. Among current users, only 4% say they are still primarily piloting; the largest group reports scaling across functions and teams, including service, IT, and merchandising. Adopters attribute improvements to customer satisfaction and personalization, followed by development velocity, operational efficiency, and employee productivity.

One figure, however, highlights a common operational gap: only 32% of organizations say they have fully defined AI success metrics and KPIs. Salesforce argues that wins are appearing before firms agree on how to measure them, and that without clear metrics it is harder to determine which investments to expand, particularly when the AI scaling effort lands on infrastructure that is already strained by fragmented customer data and systems that do not share information cleanly.

What to watch next is how quickly retailers and brands close the measurement and data-unification gap. Salesforce’s report also notes that physical retail is increasingly part of the same AI-driven journey, citing that 79% of shoppers use phones while shopping in store and that 12% ask an AI assistant for purchasing advice in the aisle. The caveat is that the findings do not specify which agentic search implementations produce the referral lift, nor how much of the growth translates into completed purchases, leaving open questions about the financial impact beyond discovery.