# Keyword Search Is Dying. Is Your Catalog Ready for AI Agents? — PayPal

## Executive summary

The e-commerce landscape is shifting from the 'search era' (keyword-based) to the 'intent era' and eventually the 'delegation era,' driven by AI agents. Traditional product catalogs, optimized for SEO and human search, are insufficient for agentic commerce. Merchants must adopt a hybrid data strategy that combines the precision of keyword matching with the breadth of semantic understanding. Successful implementation requires enriching product data with structured, high-quality content (e.g., attributes, buyer context, trust signals) rather than simply adding more unstructured text.

## Key takeaways

- The Evolution of Commerce: Shopping behavior is progressing from the Search Era (keywords) to the Intent Era (expressing needs in natural language) and finally to the Delegation Era (humans delegating tasks to agents).
- Catalog Readiness for Agents: Existing catalog specifications (syndicated for ads/SEO) were not built for AI agents. Agents require product data that is optimized for deep, semantic discovery, not just human search.
- Hybrid Search is Optimal: The ideal search mechanism combines the precision of keyword search (literal matching) with the breadth of semantic search (matching based on meaning/intent).
- Content Quality Over Quantity: Enrichment is highly beneficial, especially for merchants with thin catalogs. However, adding unstructured boilerplate text or excessive context can dilute the signal and cause agents to hallucinate.

## Technical details

- Search Paradigms: Keyword search is precise and fast but misses intent. Semantic search converts queries into vectors to match meaning (intent) but can be fuzzy. The recommended approach is a hybrid model combining both.
- Data Enrichment Patterns: Effective enrichment involves filling attributes, adding buyer context, and incorporating trust signals (e.g., review data). The approach must be tailored to the existing catalog structure.
- Agentic Commerce Metrics: Experiments showed that enriched product data significantly improved agent recommendations. Merchants with the thinnest catalogs and weak product data had the most room for improvement.

## Practical implications

- Audit existing product catalogs to determine if they are structured for agent consumption rather than human SEO.
- Implement a hybrid search architecture that balances keyword precision with semantic understanding.
- Prioritize structured data enrichment (attributes, context, trust signals) over simply increasing the volume of text.
- Focus on making the catalog discoverable for AI agents, which requires rethinking data schemas beyond traditional advertising platform specifications.

## Topics

Agentic Commerce, Semantic Search, E-commerce Data Modeling, Product Catalog Enrichment, AI Agents, PayPal Agentic Commerce Services, Agentic commerce docs

Source: https://www.youtube.com/watch?v=YXb2Wx1r_Pk