The SEO arms race is about to become obsolete—your next customer might be a bot that doesn't care about your five-star reviews.
The Summary
- Stripe co-founder John Collison argues AI agents shopping on behalf of humans will fundamentally reshape e-commerce, moving beyond targeted ads and algorithmic feeds
- Collison calls keyword search a "ridiculous" way to find products to buy, signaling the end of SEO as we know it
- Retailers will need to appeal to AI agents instead of human buyers, raising questions about whether agents can truly replicate human taste
- The payment processing giant has front-row seats to how commerce is changing—what they're seeing matters for anyone selling anything online
The Signal
Stripe processes hundreds of billions in payments annually, so when John Collison says agentic commerce represents a "sea change", he's reading real transaction data, not trend reports. The current e-commerce stack—targeted ads, SEO optimization, endless scroll—was built for human attention spans and shopping behavior. That infrastructure becomes irrelevant when your customer is an agent optimizing for parameters you can't A/B test.
The question isn't whether agents will shop for us. They already are, in limited ways. The question is what happens when they become the primary interface between consumer intent and merchant inventory. Collison's comment about keyword search being "ridiculous" cuts deeper than it sounds. Search assumes you know what you want and can describe it. Agents assume they know what you want better than you do.
"What it means when brands will have to appeal to AI agents as opposed to human buyers."
Here's where it gets interesting for anyone running an online business. The current playbook is about triggering human psychology: scarcity messaging, social proof, aspirational lifestyle imagery. An agent doesn't care about your Instagram aesthetic or that 10,000 people bought this yesterday. It cares about specs, price, delivery time, return policy, and whether the product maps to its user's stated preferences and purchasing history. That's a different game entirely.
The taste question is the unsolved problem. Can AI agents truly mimic human taste? Collison raises it because it's the hardest part. An agent can optimize for "best running shoe under $150 with good arch support." It struggles with "a dress that makes me feel like the person I want to be at this wedding." Commerce isn't just matching features to needs. It's identity, aspiration, and aesthetics—the messy human stuff that doesn't compress into clean parameters.
But agents don't need to be perfect at taste. They just need to be better than scrolling through 47 pages of search results while targeted ads follow you around the internet. That's a low bar. And every transaction they complete trains them further on your actual preferences, not the demographic bucket some ad platform thinks you're in.
The Implication
If you're building for e-commerce, start thinking about how your product looks to an API, not an Instagram feed. Structured data, clear specs, machine-readable attributes—these become your storefront. The brands that win will be the ones that make it easy for agents to understand exactly what they're selling and why it matches what the human behind the agent actually wants.
For consumers, this could mean the end of being manipulated by dark patterns and algorithmic feeds designed to maximize engagement over satisfaction. Or it could mean outsourcing taste to models trained on patterns that flatten what makes you different. The agents are coming either way. What matters is whether they're working for you or for whoever fine-tuned them.