Optimize Your Products for Agents

Everyone talks about GEO—Generative Engine Optimization—but what should you actually do about your pages? They should be "optimized," but what does that actually mean?

I kept running into this question while trying to figure out how to prepare content for AI shopping agents. The advice is everywhere, but it's frustratingly vague.

Here's the thing: no one really knows the perfect recipe yet—and I don't either. But there are a few things that sources seem to agree on, and they're worth doing regardless of whether you believe in agent e-commerce.

Why matching user language matters

LLMs are essentially solving for the most likely and appealing answers to whatever question the user asked them. And they very much like repetition and similarity. When someone asks "Will this water bottle fit in my car's cup holder?", if your content actually matches those words—"fitting" "in a car" "cup holder" (it doesn't have to be exact)—it has a much higher probability of appearing in the result.

It's a terrible simplification, but it helps you understand the game: use the words & intents people actually search for, surrounded by clear, citable chunks that directly answer their questions.

What this looks like in practice

Instead of walls of text, write in "intent modules"—small, self-contained sections that each answer one specific question shoppers have.

Break content into citable chunks: 120-250 word sections with clear headers

Match user language: Use words people actually search for, not internal product jargon

Add structured data: JSON-LD markup is an easy win—research shows it doesn't necessarily drive more impressions, but it helps retain higher quality data and gets your products into Google's carousels. Case studies consistently show better click-through rates from improved product positioning (higher traffic!)

Include quick Q&As: 2-3 FAQs that resolve common blockers like fit, returns, or setup

Simple workflow I use

When I'm optimizing any product page, I follow this pattern:

List 5-10 likely shopper questions (Who is it for? What problem does it solve? How does it fit/work?)
Rewrite existing copy into focused modules that answer each intent
Add 2-3 brief FAQs for the biggest friction points
Include basic structured data (Product, Offer, FAQPage schema)

Quick before/after

Original:

"500 ml insulated stainless bottle. Keeps drinks cold 12h. BPA-free. Hand wash. Fits most cup holders."

Rewritten module:

"The NordTrail 500 ml Insulated Bottle (NT-500) uses double-wall stainless steel to keep drinks cold for up to 12 hours when you're commuting, at the gym, or on hikes. The leak-proof twist cap rides in backpacks without spills, and the 66 mm mouth makes it easy to add ice."

What's still unclear

This field is evolving fast. No one has the perfect recipe yet, and different types of products probably need different approaches. What works for electronics might not work for clothing or consumables.

The key is starting somewhere and testing what happens. Focus on making content more helpful for humans first—if it answers real questions clearly, it's probably better for AI agents too.

Want to experiment with your content?

Try the App to see how different content structures perform, or use the Chat to explore specific questions about optimizing your product pages for AI agents.