In the days after Google rolled out AI Overviews, many merchants watched search traffic drop almost overnight. The data soon confirmed what they were seeing. According to a Pew Research Center study, users clicked a traditional search result in only 8% of searches that showed an AI summary, roughly half the rate of searches without one.
Here is the part worth sitting with: The introduction of AI Overviews was a gentle change. Everything that existed before was still there; one small region of the page became an AI answer. If a soft interface tweak can rewire traffic in days, imagine what happens when the interface itself is replaced.
Having spent years building e-commerce infrastructure for independent brands, I believe each coming change will be more fundamental than the last and that commerce is approaching its iPhone moment—a complete interface shift.
Content, products and ads are becoming the same thing.
E-commerce has long treated content, product and advertising as three separate disciplines with three separate teams and budgets. But the categories were always artificial. A shoppable product card in a social feed is simultaneously content, an ad and a transaction. The lines are dissolving.
What the open web never had was a recommendation system built around products, transactions and fulfillment—the way social platforms built one around engagement. That absence is precisely why customer acquisition costs for direct-to-consumer brands have become punishing, and why so many have retreated into retention marketing, communities and physical retail.
AI agents change this equation. When a shopping agent handles discovery, evaluation and checkout in a single conversation, the funnel collapses into one step. Content, product and ad finally converge—inside the agent’s answer.
The agent-to-agent future may be closer than it appears.
This is not speculative. Adobe Analytics reported that traffic to U.S. retail sites from generative AI sources jumped 1,200% in under a year, and during the 2025 holiday season, AI-referred retail traffic grew 693% year over year—with those shoppers converting 31% more than visitors from traditional search.
Meanwhile, agentic checkout protocols backed by major AI and payments companies are turning “an agent buys on your behalf” from a demo into infrastructure. The logical endpoint is agent-to-agent commerce: a buyer’s agent negotiating with a seller’s agent. In that world, ranking is no longer won by ad budgets and search engine optimization tricks. It is won by structured, verifiable, machine-readable context—the facts about your products that an agent can check, cite and trust.
‘Wait and see’ is the riskiest strategy.
A discipline I recommend to every founder: Be optimistic about your own execution timelines, but assume technology itself will move two to three times faster than you expect. Text-to-image AI went from research curiosity to production-grade in a few years. Nobody’s roadmap priced that in.
For merchants, the uncomfortable implication is that starting today is already late—the right time to start preparing for agentic commerce was months ago. The AI Overviews release showed that these transitions do not come with a grace period. A two-day feature rollout can permanently erase a channel you spent five years optimizing.
What can merchants do now?
1. Treat product data as your new storefront. Specifications, materials, provenance, policies and reviews need to be structured and verifiable, not buried in marketing copy. Agents buy what they can verify.
2. Measure your agentic surfaces. You track your Google rankings; start tracking whether and how AI assistants recommend you, and what they get wrong.
3. Audit your dependence on human eyeballs. Any channel that assumes a person is scrolling—paid social, SEO landing pages, email capture pop-ups—deserves a stress test against a future where an agent visits instead.
4. If you sell technology to merchants, sell the foundation, not the quick win. The question I ask about every feature is: Is this something the customer will still rely on in three years, or something we merely managed to sell them? Short-term revenue is easy to manufacture; becoming infrastructure is not.
The endgame: Production is sales, and design is brand.
Selling to businesses is not the end state. If agents come to mediate both demand and supply, most intermediaries—arbitrage sellers, dropshippers, layers of markup—can simply compress away. Factories connect to demand directly; pricing becomes dynamic and transparent.
What survives is what was always scarce: brand, taste and trust. Hence my two-line summary of the endgame: Production becomes sales, and design becomes the brand.
In my view, the iPhone’s lesson was never that phones got better. It was that a new interface quietly rewrote every industry it touched. The agent could be that interface for commerce—and the merchants who treat it as infrastructure to build on, rather than a feature to react to, could be the ones still standing when the funnel finally collapses into a conversation.
Feature image credit: Getty
By Haozhuang Dai
Find Haozhuang Dai on LinkedIn. Visit Haozhuang’s website.
