Stop A/B Testing Button Colors While AI Rewrites Your Entire Funnel.

I still see CRO roadmaps built entirely around things like button color, hero image variant, headline phrasing, the classic toolkit. All useful, historically. None of it matters even slightly to an AI shopping agent that’s evaluating your product page by parsing structured data and text content, not by looking at whether your CTA button is orange or blue.

We’re optimizing for a visitor type that’s shrinking, while a completely different visitor type, one that doesn’t have eyes, is starting to make an increasing share of purchase decisions.

The Blind Spot

Your funnel has two very different audiences now, and you’re only testing for one

A human visitor experiences your page visually. Layout, hierarchy, color, motion, all of it genuinely influences behavior, which is why classic CRO exists and has always worked. An AI agent shopping on someone’s behalf doesn’t experience any of that. It reads your page’s underlying content and structure, extracts the facts it needs, price, specs, availability, review sentiment, shipping terms, and makes a decision based on that extraction. If your key differentiators live inside an image, a fancy animated component, or vague marketing language instead of clear, parseable text, the agent effectively can’t see them.

You could win every A/B test you run on visual elements and still be functionally invisible to an increasing share of actual purchase decisions.

What “Testing For Agents” Actually Means

A different kind of optimization, running alongside the old kind

This isn’t about abandoning traditional CRO, human visitors aren’t going anywhere. It’s about adding a second, parallel discipline: making sure your page’s actual content, not just its visual design, clearly and accurately represents what makes your product the right choice, in language and structure a machine can extract cleanly.

That means specific, factual product descriptions instead of vague brand language. It means structured data (schema markup, clear specs tables, explicit pricing) instead of information buried inside a hero graphic. It means your unique value proposition needs to exist as actual parseable text somewhere on the page, not just as a vibe communicated through design.

Practical Starting Points

Where I’d actually start

• Audit your top product or landing pages by reading only the raw text and structured data, no images, no design. Would an agent understand why to choose you?
• Move key differentiators out of images and into real text and schema markup.
• Keep pricing, specs, and availability explicit and structured, not buried in prose or requiring a click to reveal.
• Add “testing for agent comprehension” as a real line item in your CRO roadmap, not an afterthought bolted onto the visual testing calendar.

Quick Answers

A few direct questions, answered directly

Do AI shopping agents care about website design?
No, AI agents evaluate products based on parseable content and structured data such as price, specifications, and reviews, not visual design elements like color or layout.

Should companies still do traditional A/B testing?
Yes, for human visitors traditional CRO remains valuable. The recommendation is to add agent-focused optimization alongside it, not replace it.

How do I optimize a page for AI shopping agents?
Ensure key product information and differentiators exist as clear, structured, parseable text and schema markup, rather than being communicated only through images or design.