An Agentic Commerce Reality Check

Source: Andrew Lipsman (guest post); The AI Ad Economy (Substack) · 2026-04-23

I’ve previously described agentic commerce as a “collective hallucination”—a not-so-popular characterization among the pro-agentic crowd.

In late fall 2025, all three converged in retail and ecommerce. Retailers were under growing pressure to squeeze more revenue from retail media networks, even as performance gains became harder to sustain. At the same time, the surging adoption of AI search made the idea of “agentic commerce” sound not just plausible but inevitable—a narrative too compelling for an anxious industry to resist. The Fall conference season echo chamber—amplified on LinkedIn—converted speculation into conviction almost overnight.

The sentiment soon crystallized as every major consulting firm weighed in with its prognosis of retailers’ existential risk:

“Agentic commerce will be the most disruptive retail shift since e-commerce. 60% of consumers expect to use AI agents for shopping within the next 12 months. The real threat comes from cross-platform ‘super agents’ that sit between retailers and consumers. For unprepared retailers, the risk is up to 500 bps EBIT erosion from margin compression, traffic dispersion, and brand commoditization.”

“The advancements in agentic shopping present substantial challenges for traditional retailers, primarily through risks associated with disintermediation... The growth of zero-click search and agent-driven interactions is eroding direct traffic—along with the retailer’s ability to observe, influence, and understand consumer behavior at scale... AI agents’ emphasis on utility weakens traditional brand loyalty, limits cross-selling opportunities, and puts retail media revenues under pressure, all while intensifying the race to the bottom on price and delivery speed.”

“Agentic commerce also threatens to bring about a decline in traditional revenue streams, particularly from advertising. Retail media networks, which rely on ad-based models, could face challenges as consumers shift toward agent-driven experiences that bypass traditional ad channels.”

“Of the three types of AI agents – third-party ‘objective’ agents (e.g. ChatGPT, Perplexity), on-site retailer agents, and off-site retailer agents – third-party agents threaten to disintermediate retailers most.”

The result was a near-industry-wide belief that agentic commerce would rewrite the rules of retail within months. An ongoing array of headline-grabbing charts and statistics have only added fuel to the fire.

But few of these stats stand up to scrutiny.

Throughout the 2025 holiday season, Adobe reported on various ecommerce sales and shopping trends. Garnering the most headlines this year wasn’t the increase in ecommerce spending but the dramatic rise in Gen AI referral traffic: “Generative AI tools drove a 693.4% increase in traffic to retail sites”.

I received pitches from Adobe’s PR firm throughout the season highlighting the growth rates. On at least one occasion, I asked what the basis of the growth rate was but never received the relevant supporting data. While ~700% growth might sound impressive, year-over-year growth rates like that tend to only occur from a very small base.

At the close of the season, my curiosity led me to check another reliable source for web traffic: Sensor Tower. This data provided the clarity needed to put this trend in context, revealing that Gen AI referral numbers increased from 0.1% of traffic to 0.6% between Holiday 2024 and Holiday 2025.

The overall growth rate in this traffic jumped ~200%, an impressive number to be sure. (Not quite as high as Adobe’s estimate, but directionally similar—the difference likely explained by Adobe’s footprint of ecommerce sites not including Amazon and other large retailers.) And yet, the total Gen AI referral still barely registers in the overall composition of ecommerce traffic.

Adobe’s headlines weren’t inaccurate, but they did lack important context.

One school of thought says that agentic capabilities will shift shopping to LLMs, drive ecommerce traffic to zero, and vaporize on-site retail media revenue in the process. And this will all happen by the end of the decade!

The directive is clear: retailers should forsake current initiatives—such as RMN buildouts—to adapt to this zero-traffic future.

I decided to consult Emarketer, a reputable third-party source of forecast data, which has projections for retail/commerce media and agentic commerce through 2029. According to those estimates, “agentic commerce” (“AI platform on-site retail ecommerce sales”, per Emarketer) will reach $28 billion by that date. Meanwhile, commerce media ad spend will increase by $48 billion from its 2025 total. Thus, if retailers allocate resources based on the revenue potential, then retail media represents a bigger opportunity to pursue.

If we apply typical margins of 10% for ecommerce and 60% for retail media, the gap between the two opportunities gets more dramatic. Under this scenario, commerce media would produce $29 billion in operating margin vs. $2.8 billion for ecommerce occurring on AI platforms. (Feel free to use different margins assumptions, but the disparity will still be significant.)

There’s no way to predict this future with certainty, but retailers can make decisions rooted in rationality over hyperbole. And when you do the math, the answer should be clear.

A chart began circulating on Linkedin late last year suggesting that ChatGPT’s meteoric rise as a commerce channel made it the #2 ecommerce site in web traffic after Amazon.

A back-of-the-napkin calculation used OpenAI’s self-reported stats of 2.5 billion prompts per day and 2.1% pertaining to purchasable products to get 52.2 million queries. This was then compared to Euromonitor’s US daily visits to Amazon, Walmart, Ebay and other leading retailers. Only Amazon topped 52.2 million.

It made for a great headline—if only it were close to accurate.