Agentic search up 200%: measure your AI referrals

Salesforce published its Fourth Edition State of Commerce report on July 28, and one number is getting quoted everywhere: agentic search as the first step in a purchase journey grew 200% year over year. Before you rebuild your storefront around that, understand what was measured — and then go check whether it's happening to you, because most of the businesses we talk to cannot answer that question from their own analytics.

What actually happened

Per Salesforce's writeup, the report combines three data sets: a survey of 3,450 commerce professionals across 20 countries (April 10–June 4, 2026), a consumer survey of 4,689 shoppers across eight countries (May 12–18), and platform behavioral data from more than 1.5 billion shoppers across 37 countries covering Q1 2024 through Q1 2026.

The behavioral half is the part worth your attention, because it's observed activity rather than self-report. Traffic referred from AI chats grew between 150% and 428% year over year in every quarter measured, while total traffic grew in the single to low double digits. Over August 2025 to May 2026, discovery through brand-owned properties fell 7% and traditional search fell 15%, while discovery through AI assistants, social AI, and delivery apps rose 38%.

The survey half is softer. 86% of commerce leaders say LLMs will be essential to product discovery within a year. Only 28% currently use agentic AI, with 44% planning to inside six months. And the finding that undercuts the rest: just 32% of organizations have fully defined AI success metrics. Two-thirds are reporting wins without having agreed what winning is.

Worth naming: Salesforce sells Agentforce Commerce. Vendor research that concludes you urgently need the vendor's category is directionally useful and never a substitute for your own numbers.

Why AI referral traffic matters for your business

Here's the practical problem. Percentage growth off a small base is still a small base, and the report doesn't publish AI chat referrals as a share of total sessions. So the honest question isn't "is agentic search growing 200%" — it's "what is it for my catalog, this month." That's a two-hour job, not a replatform.

Go look:

1. Split AI assistants out of "Direct" and "Referral." In GA4, build a custom channel group matching referrer hostnames — chatgpt.com, perplexity.ai, gemini.google.com, copilot.microsoft.com, claude.ai. Better: read raw Referer headers off your edge logs, because a chunk of assistant traffic arrives with no referrer at all and lands in Direct.
2. Log the agents separately from the humans. Requests from OAI-SearchBot, PerplexityBot, ClaudeBot, and Google-Extended in your access logs tell you whether crawlers can even reach your product pages. If your PDP renders client-side behind a JS bundle, they're reading an empty div.
3. Make the catalog machine-legible. Valid schema.org/Product with offers, price, priceCurrency, and availability on every PDP, and a real product feed. Assistants answer from structured data — the merchandising copy you paid for is not what gets parsed.
4. Define the metric before the project. Sessions from AI sources, their conversion rate, and revenue per session, tracked monthly. If it's 0.4% of revenue, you have a monitoring task. If it's 6%, you have a channel.

We build for the second case and instrument for the first. The order matters.

Sources: Salesforce — Shopping's New First Step: Agentic Search Grows 200%.