The marketing playbook you've refined for decades is about to become a liability.
"Only 3 left in stock!" That message drives human urgency. It also tanks AI recommendations. Research shows scarcity messaging reduces AI recommendations by 24-46% because machines prioritize reliability over FOMO.
Welcome to the era of agent psychology.
The Shift No One's Talking About
Most AI commerce conversations focus on one question: "Are you visible to AI search?" The entire field of Answer Engine Optimization (AEO) has emerged around getting mentioned.
But Geoff Gibbins and his team at Human Machines asked a different question: "When AI finds you, will it recommend you?"
The distinction matters. Walmart has ChatGPT Instant Checkout integration and access to 800M+ weekly ChatGPT users. Their "Agent Psychology" score? 20 out of 100. Access without optimization is opportunity wasted.
AI Agents Are Weird in Specific, Measurable Ways
Human Machines ran an analysis of 134 major US brands using their Reconnix tool, testing how ChatGPT, Gemini, and Claude interpret web content. What they found challenges everything marketers assume about persuasion:
Position bias creates winner-take-all dynamics. In Microsoft's 2025 Magentic Marketplace study, Claude Sonnet 4.5 accepted the first proposal it received about 90% of the time. Being the agent's first choice can mean category dominance. Being second can mean irrelevance.
Different AI models have different spatial preferences. GPT-4.1 consistently favors products positioned on the left side of pages. Claude favors the center. Gemini prefers the right column. Your page layout now has three audiences with conflicting preferences.
Irony and humor register as negative signals. Oatly's clever, unconventional brand voice that resonates with humans? AI systems interpret it as unreliability. Their Agent Psychology score: 21.
Trust signals trapped in images are invisible. Star ratings, review counts, and certifications rendered as graphics rather than HTML text don't exist to AI crawlers.
The Four Dimensions That Matter
Human Machines' Reconnix framework breaks AI commerce readiness into four categories:
Discoverability (30-35%): Can AI actually find your content? Server-side rendering, structured data markup, and crawler access determine whether you exist to AI systems at all.
External Authority (35-40%): What does the rest of the internet say about you? Third-party reviews, earned media, community presence on Reddit, and reference sources like Wikipedia.
Agent Psychology (25%): If AI finds you, will it recommend you? This measures attribute strength, social proof in readable formats, platform signals, and critically, pressure-free messaging.
Commerce Readiness (0-10%): Can AI actually transact with you? Product feeds, payment integration, conversational commerce capabilities.
Who's Winning (and Why)
The top performers share common traits: Under Armour, Casper, Everlane, and REI all use server-side rendering for immediate content accessibility. They have comprehensive schema markup with 85-99% coverage. They emphasize reliability messaging over urgency.
Digital-native brands show a structural advantage, averaging 52.3 versus 36.8 for traditional retail. That 42% gap reflects infrastructure decisions, not marketing budgets.
The companies struggling? Those with JavaScript-rendered content AI crawlers can't see. Those using scarcity and urgency tactics that worked on humans. Those with trust signals locked in images.
The Permanent Capability You Need
Here's what makes this genuinely challenging: human psychology remains relatively stable. Agent psychology changes with every model release.
The irrationalities of GPT-4 differ from GPT-4.1. Claude's biases differ from Gemini's. And both will shift again in the next training run. This isn't a one-time optimization project. It's a permanent organizational capability.
Companies that treat AI commerce readiness as a campaign will find themselves perpetually behind. Companies that build ongoing testing and optimization into their operations will compound their advantage.
What You Can Do This Week
Run your brand through Reconnix. Geoff's team at Human Machines built a tool that tests how ChatGPT, Gemini, and Claude actually perceive your website. Get your score at Reconnix. If you want to dive deeper on your results or benchmark against competitors, reach out to Geoff directly at geoff@human-machines.com.
Audit your urgency messaging. Search your codebase for "limited time," "only X left," "ends soon," "last chance." Each instance may be actively reducing AI recommendations by 24-46%.
Check whether your trust signals are visible. Are your star ratings, review counts, and certifications in HTML text or trapped in images? AI can't read your 5-star graphic.
Jason Hauer is CEO and Founder of HauerX Holdings, backing and building AI-native companies that deliver compounding growth for the now. Subscribe to the Tuesday Growth Brief at hauerx.com.
