Most merchants treat product titles as labels and descriptions as marketing copy. Both assumptions made sense when the end reader was a human scanning a page.
AI agents don't scan. They retrieve, parse, and match. The structure of your text is part of the signal, not just a formatting preference.
What makes a title agent-readable vs. SEO-optimized
SEO-optimized:
"Best Lightweight Face Moisturizer for Sensitive Skin — Fragrance-Free, 1.7 fl oz"
Agent-readable:
"Cetaphil Moisturizing Lotion, sensitive and dry skin, fragrance-free, 1.7 fl oz, daily face and body use"
The second follows the structure agents use to extract product data: Brand + Product type + Key attributes (comma-separated) + Use case.
Prose vs. structured descriptions
Prose descriptions require agents to parse natural language before they can extract and validate attributes.
Before (prose): "A rich yet lightweight moisturizer formulated for those with sensitive, easily irritated skin. Dermatologist-tested and free from harsh fragrances."
After (structured): A brief editorial sentence followed by a spec table with Skin type, Key claims, Size, Suitable for, and Not recommended for.
Field consistency: the catalog-wide problem
When an agent is evaluating multiple products in a category, it normalizes attributes across listings to build a comparison set. If your moisturizer calls it "Skin type: Sensitive" but another product calls it "For: All skin types, including sensitive," the agent may not recognize these as compatible data points.